Introduction
The main aim of this book is to establish the regular reflexes of the syllabic liquids *rÌ¥ and *lÌ¥ in all Ancient Greek dialects, including Mycenaean. These sounds were inherited by Proto-Greek from Proto-Indo-European as allophones of /r/ and /l/ in a number of phonological environments.1 All first millennium Greek dialects have lost *rÌ¥ and *lÌ¥, as did most other Indo-European languages upon their first attestation. However, Proto-Greek must have retained them because the alphabetic dialects show various different reflexes of *rÌ¥ and *lÌ¥. For example, the Proto-Greek thematic aorist *amrÌ¥t-e/oâ âto miss, failâ is continued in Ionic-Attic as ἥμαÏÏον, inf. á¼Î¼Î±ÏÏεá¿Î½, but in Lesbian as inf. αμβÏοÏην (epigraphically)2 and ind. á¼Î¼Î²ÏοÏε (Sapph.). Proto-Greek *trÌ¥pedi̯a âtableâ is reflected as to-pe-za in Mycenaean, but as ÏÏá½±Ïεζα in alphabetic Greek from Homer onwards.
Questions that a historical linguist may ask regarding such forms are: what conditioned the difference between the reflexes âαÏâ (á¼Î¼Î±ÏÏεá¿Î½) and âÏαâ (ÏÏá½±Ïεζα)? What does the Mycenaean form to-pe-za represent phonologically and phonetically? Why does Lesbian have a reflex âÏοâ in á¼Î¼Î²ÏοÏε, but âαÏâ or âÏαâ in several other forms? Is the o-reflex, in those dialects where we find it, subject to phonological conditioning or is it found across the board?
The bibliography on the syllabic liquids in Greek is large, and not every previous treatment of the topic will receive equal attention in this book. In section 1.1, where I discuss a selection of previous scholarship, the main aim is to illustrate the different issues that are at stake. Once these issues have been presented, the scope of this investigation will be delimited more precisely.
1.1 The Problem and Its Relevance
Determining the regular reflexes of *r̥ and *l̥ is not just an issue of Greek historical phonology. The problem is intimately connected with two other, much-debated questions that are of considerable importance for reconstructions of the prehistory of the Greek language and literary traditions. First, how and when did the four main dialect groups of alphabetic Greek originate? And secondly, how and when did the artificial language of the epic tradition, in the form familiar to us from Homer onwards, come into being? Once the regular reflexes of the syllabic liquids and the chronology of their developments have been established, more definite answers to these questions may be given.
1.1.1 A Concise Summary of Some Previous Accounts
Although scholars disagree on many smaller issues, in essence there have been three basic views on the development of the syllabic liquids. I will associate these views with the names of Ruijgh, Tichy, and Heubeck.
C.J. Ruijgh has written about the syllabic liquids in Greek for almost his entire scholarly career.3 Since he usually presents his views and their consequences with great clarity, his work is a good place to start. Ruijghâs two main contentions are:
-
The syllabic liquids were eliminated from all Greek dialects already in the mid-second millennium. This resulted in a split between dialects with o-vocalism (Aeolic, âAchaeanâ4) and dialects with a-vocalism (Ionic-Attic, West Greek).
-
The metrical behavior of certain Homeric formulae containing a reflex of *r̥ proves that epic composition in hexameter verse (more or less in the form known from Homer) existed as early as the mid-second millennium.
Concerning point 1. it is traditionally accepted that a regular o-colored reflex of the syllabic liquids is found only in the Aeolic dialects (Lesbian, Thessalian, Boeotian) and in Arcado-Cyprian.5 From the viewpoint of Classical Ionic-Attic, this reflex was considered so characteristic that Aeolic and Arcado-Cyprian were occasionally lumped together, in the first half of the previous century, as a special subgroup. After the decipherment of Linear B, however, most scholars have come to agree that the fundamental division is between what Risch (1955) called North Greek and South Greek.6 The two most important isoglosses separating these two groups are the South Greek assibilation *t(h)i > si and the South Greek development of intervocalic *ât(h)i̯â through *âtsâ to âssâ (later > Ion.-Att. and Arc. âÏâ). The phonologically more conservative North Greek dialects initially retained t(h)i and *ts.7
In discussions of early Greek subgrouping the reflexes of *rÌ¥ have played an important role, especially when the decipherment of Linear B seemed to prove an early date for its vocalization. Mycenaean forms like to-pe-za and qe-to-ro-po-pi, which derive from PGr. *trÌ¥-ped-i̯a and *kwetrÌ¥-pod-phi, are usually thought to represent /torpeddja/8 and /kwetropopphi/, respectively. These examples seem to prove that the vocalization had been accomplished already in the early 14th c. BCE (the earliest attestations of Linear B) in the âAchaeanâ dialects of South Greek, and perhaps even earlier.
A much-cited argument in this connection is the development of an epenthetic âdâ between a coda nasal and an onset liquid.9 This phenomenon is attested already in Mycenaean10 and also in the Homeric form á¼Î½Î´ÏοÏá¿Ïα âvigorâ, which is usually taken to reflect PGr. *anrÌ¥tÄtâ. Since the insertion of âdâ in á¼Î½Î´ÏοÏá¿Ïα presupposes the vocalization of *rÌ¥ to âroâ, and since the same reflex appears to be found in Myc. qe-to-ro-po-pi, most scholars have concluded that both developments, the vocalization of *rÌ¥ and d-epenthesis, took place in this word prior to the attestation of Linear B. By extension, it was assumed that the other dialects vocalized *rÌ¥ (and *lÌ¥) around the same time, even if these dialects are first attested at a much later date than Mycenaean.11
Starting from these assumptions, Ruijgh concludes that the developments represented in Table 1 took place in the mid-second millennium, resulting in a split into four dialect groups.12
Table 1
The mid-2nd millennium split into 4 dialect groups, according to Ruijgh
|
South Greek |
*t(h)i > si |
Achaean |
*rÌ¥ > âorâ, âro- |
|
*ât(h)i̯â > *âtsâ > âssâ |
Ionic-Attic |
*rÌ¥ > âarâ, âra- |
|
|
North Greek |
*t(h)i retained |
Aeolic |
*rÌ¥ > âorâ, âro- |
|
*ât(h)i̯â > *âtsâ |
West Greek |
*rÌ¥ > âarâ, âra- |
Note that the argument for a mid-second millennium split into four dialect groups depends also on morphological criteria, but the outcome of *rÌ¥ is the only phonological criterion used in this connection. There are no other phonological developments that are demonstrably early and where the first millennium dialect groups have different reflexes.13 It does not come as a surprise, then, that alternatives to Ruijghâs scenario have been put forward. Risch (1955) maintained that there were no significant differences between âAchaeanâ and Proto-Ionic in the Mycenaean period, and denies that the reflexes of *rÌ¥ can be used as a reliable criterion.14 Heubeck (1972) argued that *rÌ¥ was preserved in Mycenaean, and he was followed by GarcÃa Ramón (1975), who claimed in addition that Proto-Aeolic retained *rÌ¥ until a relatively late date.
Let us now turn to the second issue: the prehistory of Epic Greek and the hexameter. The history of scholarship on this question in the second half of the 20th century is well summarized by Hajnal (2003). The debate was initiated by Mühlestein (1958) in an article about Mycenaean names starting with a-noâ. He interpreted this as representing the lexical element /anor-/ < *anrÌ¥â âmanâ, corresponding to Class. á¼Î½Î´Ïοâ < *anroâ. Moreover, he connected the Mycenaean names with a long-standing metrical problem from Homeric Greek. The verse-final coordinated noun phrase á¼Î½Î´ÏοÏá¿Ïα καὶ ἥβην âvigor and youthâ and the noun-epithet formula á¼Î½Ï αλίῳ á¼Î½Î´ÏεÏÏόνÏá¿ âto man-slaying Enualiosâ, which both seem to be old elements of epic diction, are unmetrical as they appear in our Homeric text. This remains so if we replace the morphologically opaque form á¼Î½Î´ÏεÏÏόνÏá¿ by a pre-form with *á¼Î½Î´Ïοâ. Moreover, using a formulaic phrase like á¼ÏÏá½·Î´Î¿Ï á¼Î¼ÏιβÏá½¹ÏÎ·Ï âman-covering shieldâ in the epic hexameter requires that plosive plus liquid (PL) onsets can be realized as tautosyllabic. Of course, muta cum liquida scansion is a well-known license in many varieties of Greek poetry, but in Homer this phenomenon is relatively rare, and highly uncommon with word-internal PL-clusters.15 These metrical irregularities disappear if *rÌ¥ is substituted for its Homeric outcome âÏοâ, i.e. in the reconstructed pre-forms *anrÌ¥tÄta, *anrÌ¥kwhontÄi, and *amphimrÌ¥tÄs. Hence, it is attractive to assume that the phrases in question were coined before *rÌ¥ was eliminated from the dialect in which they were composed. Now, if it is true that the change *rÌ¥ > âorâ, âroâ had been completed in Mycenaean already before our attestations of Linear B, as many scholars assume (see above), and if we also believe that pre-forms of á¼Î¼ÏιβÏá½¹ÏÎ·Ï and á¼Î½Î´ÏοÏá¿Ïα (with their reflex âÏοâ) entered the epic tradition from this direct ancestor of Mycenaean, it would follow that the formulae in question were coined by poets speaking this pre-form of Mycenaean, approximately in the mid-second millennium BCE.16
This account of á¼Î½Î´ÏοÏá¿Ïα and related forms was widely shared in the 1980âs and early 1990âs, but in the meantime, it had also become the topic of a controversy initiated by Tichy (1981).17 Tichyâs main objection to the account just sketched was that the unchanging existence of the Homeric hexameter for such a long period is a premise that cannot be relied upon. She argued, instead, that phrases like á¼ÏÏá½·Î´Î¿Ï á¼Î¼ÏιβÏá½¹ÏÎ·Ï and á¼Î½Ï αλίῳ á¼Î½Î´ÏεÏÏόνÏá¿ could be relatively recent creations. Moreover, she maintained that á¼Î½Î´ÏοÏá¿Ïα καὶ ἥβην does not provide evidence for a phonological pre-stage, but for an older shape of the verse form: she explains it by taking recourse to the proto-hexameter framework proposed by Berg (1978). In Tichyâs view, the aberrant Homeric scansion of the form á¼Î½Î´ÏοÏá¿Ïα was regular at a pre-stage of epic verse when a trochaic fourth foot was still allowed. This scenario, or at least its possibility, has gained an increasing number of proponents among Indo-European scholars.18 Another point of criticism directed at Ruijghâs views has been that the preservation of metrically irregular formulae over a period of seven centuries is implausible.19 Finally, it has been claimed that the formula á¼Î½Î´ÏοÏá¿Ïα καὶ ἥβην, in its Homeric form, cannot have existed in (preâ)Mycenaean epic because the conjunction καί is unattested in Mycenaean.20
The idea of an early split into a-coloring and o-coloring dialects has also been challenged in various different ways. First of all, the Mycenaean situation cannot be automatically projected onto the other dialect groups with o-colored reflexes: there is no cogent reason to assume a development shared by Mycenaean with Proto-Aeolic. A fortiori, we must remain agnostic about the date of vocalization of *rÌ¥ and *lÌ¥ in the other non-âAchaeanâ dialect groups.21 Moreover, the assumption that an o-vocalic reflex of *rÌ¥ and *lÌ¥ was the only regular treatment in Aeolic and âAchaeanâ has occasionally been challenged, most notably by Morpurgo Davies (1968), and more recently by Thompson (2010).22 Finally, Heubeck (1972) has argued that Mycenaean, as attested in the Linear B tablets, even preserves *rÌ¥. He proposed that the epic language and its meter first originated in the early Dark Ages, when stories about the âheroicâ age of the Mycenaeans started to be told.
To conclude this introductory discussion, there is still no consensus about the following points:
-
The exact reflexes of *r̥ in Aeolic and Arcado-Cyprian;
-
The date of its vocalization in the various dialect groups;
-
The origin of metrically aberrant forms with âÏοâ in Homer.23
Regarding the last two points there are three main positions. Scholars like Ruijgh argue for an early vocalization of *rÌ¥ in all dialect groups, and think that certain metrically anomalous forms in Homeric Greek were adopted by the epic tradition at this early time in a form with *rÌ¥. Tichy and her followers agree about the early date of the vocalization of *rÌ¥, but for them the idea of reflexes of *rÌ¥ in Homeric meter is anathema (cf. section 1.5.3). Finally, scholars like Heubeck consider a late vocalization of *rÌ¥ possible and consequently have less problems in viewing metrically aberrant verses in Homer as preserving traces of *rÌ¥. Concerning the date of vocalization, the mainstream view still sides with Ruijgh and Tichy,24 but as we will see the arguments on which this view is based are not strong. As for the possibility to distinguish metrical traces of *rÌ¥ in Homeric meter, many scholars these days have yielded to Tichyâs arguments against this.25 In chapters 6 and 7, I will plead for a reappraisal of the views held by Wathelet and Heubeck, and reinforce their case with new arguments.
1.1.2 Research Questions and Hypotheses
The attempt to disentangle this web of questions and proposed solutions must start with a thorough investigation of the evidence for each dialect, as attested in epigraphic material, in glosses and grammarians, and in literary sources. The first main goal of this book is, therefore, to review the entire evidence for *r̥ and *l̥ and to establish the regular development(s) per individual dialect group. An evaluation of the etymological evidence for *r̥ in Mycenaean and the alphabetic dialects apart from Ionic-Attic will be given in chapters 2 and 3, respectively. The much more extensive evidence for *r̥ from (literary) Ionic-Attic, including the oldest attestations in Epic Greek, is the subject of chapters 4 to 9. The development of *l̥ in all dialects is discussed separately in chapter 10, because *r̥ and *l̥ may have had different reflexes.
The main focus of my attention will be on the regular place of the anaptyctic vowel. In this respect, the present work differs from most previous treatments of the problem. Shorthand formulations like âPGr. *rÌ¥ > Ion.-Att. αÏ/Ïα, Myc. or/roâ are commonplace in the scholarly literature. However, if such statements are taken at face value, the assumed variation would violate the principle of Ausnahmslosigkeit. Since sound changes normally do not have a dual outcome, we must ask, for each individual Greek dialect: was the regular reflex âorâ, âroâ, âarâ, or âra-? The evidence for various dialects within the West Greek group is rather limited, but for the two dialect groups with an o-colored reflex we will reach a remarkable conclusion: the regular treatment in Aeolic dialects is *rÌ¥ > âÏοâ (chapter 3), but such a development can be excluded for Mycenaean (chapter 2).
The most complicated question concerns the regular outcome of *rÌ¥ in Ionic-Attic: was it âαÏâ or âÏα-? The existence of pairs like κÏαδίη ~ καÏδία and κÏαÏεÏá½¹Ï ~ καÏÏεÏá½¹Ï forms a long-standing problem to which various solutions have been suggested since the late nineteenth century. As we will see in section 1.4, none of these attempts has been particularly successful. Therefore, many scholars have resigned to the view that the original distribution cannot be fully recovered. At the same time, it is still widely believed that *rÌ¥ > Ïα was the regular development in Ionic-Atticâin spite of various unresolved problems.26
In my view, this conclusion is unwarranted, and the problematic âdouble reflexâ in Ionic-Attic must be tackled from a completely different angle. I posit a regular development *rÌ¥ > âαÏâ in Proto-Ionic (i.e. the latest common ancestor of Attic and all varieties of Ionic), and propose to explain a considerable number of instances of âÏαâ by a development taking place in Epic Greek, which affected those forms with *rÌ¥ that were retained longer within the epic tradition. I will briefly introduce the reasons for proposing such a scenario in section 1.5, and elaborate the details in chapters 6 to 8. Thus, the second objective of this book is to make explicit the various mechanisms by which forms with an original *rÌ¥ were treated in Epic Greek.
As we have seen, the reflexes of *rÌ¥ have played an important role in previous discussions about the genesis of the four main Greek dialect groups. The reader may have noticed my skepticism concerning the alleged mid-second millennium date of the vocalization. Indeed, on the basis of the new proposals made in this book, the value of the vocalization of *rÌ¥ as an isogloss must be reconsidered.27 This is the third main objective of this book. An important realization is that Aeolic and âAchaeanâ have different outcomes of *rÌ¥, in spite of the fact that they appear to share an o-colored reflex. This conclusion deprives the idea of an early vocalization of all justification: there is no longer any reason to view these two dialectal developments as part of the same isogloss.28 For reasons that will become clear later, I think that the vocalization of *rÌ¥ must be pushed forward in time as far as possible towards our first attestations.29 In this context, a particularly important question is whether Mycenaean still preserves *rÌ¥, as Heubeck (1972) argued. This thesis has been widely criticized and, as we will see in chapter 2, the issue is indeed difficult to resolve on the basis of the Mycenaean evidence alone. However, as will become clear from the scenario proposed here for the development of *rÌ¥ in Epic Greek, there are various aberrant word-forms and/or scansions in Homer that probably entered the tradition in a shape with *rÌ¥ in a Mycenaean context. In my view, such forms make the retention of *rÌ¥ in the palatial period, and the existence of a precursor of the epic tradition at that stage, likely.
Before examining previous proposals to solve to the vexed issue of the âdouble reflexâ in section 1.4, I will first of all delimit the phonological environments where the Greek dialects did not diverge in their treatment of *rÌ¥ and *lÌ¥. These environments with a Pan-Greek or Proto-Greek vocalization to /ar, al/ (or /Ér, Él/) are discussed in section 1.2, and will only play a marginal role in the remainder of this book. After that, various issues related to o-colored and alleged u-colored reflexes of *rÌ¥ will be treated in section 1.3: in which dialects do we find o-vocalism, and under which conditions? Is there any evidence for a u-colored reflex in labial environments? And, finally: is there any connection between the o-colored reflex of the syllabic liquids and that of the syllabic nasals?
1.2 Environments with a Common Greek or Proto-Greek Reflex αÏ, αλ
In Proto-Indo-European, *r̥ and *l̥ were allophones of /r/ and /l/, occurring whenever these phonemes served as a syllabic nucleus. This would usually be a consequence of ablaut, which left /r/ or /l/ between two consonants (including laryngeals), or at word end after a consonant. Examples are:
-
*CLÌ¥C: ÏÏá½³ÏÏ âto turnâ < *trekw-oH, aor. á¼ÏÏαÏον < *e-trÌ¥kw-om
-
*CLÌ¥HC: βοÏá½± âfoodâ < *gworh3-éh2, βÏÏÏόν âmeatâ < *gwrÌ¥h3-tóm (root PIE *gwerh3â)
-
*CLÌ¥HV: á¼Ïήν âlambâ < *u̯rÌ¥h1-Än, ÏολύÏÏÎ·Î½Î¿Ï ârich in lambsâ < *polh1u-u̯rÌ¥h1n-os
-
*âCLÌ¥: á¼¦Î¼Î±Ï âdayâ < *h2eh1-mrÌ¥, Myc. a-mo-ra-ma /ÄmÅr-ÄmrÌ¥/ (1st CM *h2eh1-mÅr).
It is sometimes assumed that *rÌ¥ and *lÌ¥ could occur as a consequence of Sieversâ Law, in words where a suffix like *âroâ or *âloâ followed a heavy syllable, e.g. á¼Î³Îºá½±Î»Î· âbent armâ < PIE *h2énk-leh2, á½Î¼ÏÎ±Î»á½¹Ï ânavelâ < PIE *h3nbh-lós.30 However, it remains uncertain whether Sieversâ Law was really a productive phonological rule at any stage of the prehistory of Greek.31 Moreover, even if Sieversâ Law was operative, it is not quite clear whether it makes sense to distinguish e.g. *h3nbhl̥ós as an intermediate stage, rather than assuming a direct vowel anaptyxis (PGr. *omphlós > *omphÉlós) that was at some point phonemicized. In other words, in such examples *lÌ¥ is nothing more than a notation indicating that anaptyxis took place. The same point applies to alleged cases of Lindemanâs Law in Greek, such as nom.-acc. sg. κάÏη âheadâ < *kÉrÄ < PGr. *krÄ. I will not be dealing structurally with alleged cases of prevocalic *rÌ¥ or *lÌ¥.
While *rÌ¥ and *lÌ¥ were originally allophonic variants of /r/ and /l/, they may have functioned as distinct phonemes (be it marginal ones) at certain stages of the prehistory of Greek. In Proto-Greek, the loss of intervocalic laryngeals led to a phonemicization of the glides *i̯ and *u̯, as opposed to the vowels i and u. Thus, in PIE *medhioâ > PGr. *methi̯oâ âmiddleâ, the sequence *âi̯oâ had become phonologically distinct from e.g. the suffix PGr. *âi(i̯)oâ < PIE *âiH-oâ (continued as Gr. âιοÏ). At this time, a marginal phonemic difference between consonantal /r/ /l/ and syllabic /rÌ¥/ /lÌ¥/ may also have come into existence: the sequence âurâ between two consonants (e.g. in the borrowing Ïá½»ÏÎ³Î¿Ï âfortificationâ) may have been realized differently from âu̯rÌ¥â in the same position (e.g. in PGr. *tu̯rÌ¥kes âpieces of meatâ > Ïá½±ÏκεÏ). However, it is also possible that Ïá½»ÏÎ³Î¿Ï was borrowed into Greek only after the syllabic liquid in *âu̯rÌ¥â had been eliminated. In any case, instances like Ïá½»ÏÎ³Î¿Ï versus *tu̯rÌ¥kâ would have remained marginal.
In other cases, syllabic liquids were eliminated early on (but after PIE) by conditioned phonological developments. This certainly included the environments PIE *CRHV and *CRHC.32 An early vocalization has also been envisaged for three other environments: word-initial and word-final position (cf. Schwyzer 1939: 342), and the positions before a glide (*CLi̯/u̯Vâ) and a nasal (*CLNVâ). I will now discuss these environments in succession.
1.2.1 PIE *CRHV and *CRHC
In PIE sequences of the structure *CRHV, an anaptyctic vowel had developed in Proto-Greek before the sonorant: *CÉRHV. When the laryngeals were eliminated, the anaptyctic vowels were phonemicized, meaning that the syllabic liquid was âvocalizedâ: *CÉRV. This subsequently yielded Greek /CaRV-/, e.g.:
-
βαÏá½»Ï âheavyâ < *gwrÌ¥H-uâ (cf. Ved. gurú-, Goth. kaurus, etc.)
-
Ïá½±ÏÎ¿Ï âbeforeâ < *prÌ¥Hos (cf. Ved. puráḥ âin frontâ)
-
á¼Ïήν âlambâ < *u̯rÌ¥h1-Än (cf. Ved. úranâ f. âid.â)
-
ÏÎ±Î½Î±á½¹Ï âthinâ < *tnÌ¥h2-eu̯oâ.33
Two points deserve attention. First, the development *CRHV > *CÉRV shows that all Greek dialects developed an anaptyctic vowel before the liquid or nasal, and secondly, the development is identical for liquids and nasals. This contrasts with the development in most environments not involving laryngeals, in which case an anaptyctic vowel developed after the liquid in some dialects (yielding âÏοâ or âÏαâ), but before the liquid in others (yielding âοÏâ or âαÏâ). This means that the development *CRHV > *CÉRV preceded the vocalization of the syllabic liquids (and nasals) before consonants other than laryngeals.
It is likely that the anaptyctic shwa in *CÉRHV merged with /a/, the reflex of interconsonantal *h2 and *h2e, at an early stage. The main argument for this claim is that all Greek dialects appear to have this reflex.34 Some scholars have proposed a special development *CRHV > *CoRV for Lesbian, but the evidence consists of just two forms in Alcaeus:
-
ÏόμονÏÎµÏ âcuttingâ, corresponding to Ion. ÏαμόνÏÎµÏ < *tmÌ¥h1-ont-es;
-
ÏόλαιÏι corresponding to Ion. ÏαλῶÏι (3pl. ind. pres. of ÏÎ±Î»á½±Ï âto release; slackenâ).
Concerning ÏόμονÏεÏ, I agree with Francis (1974: 23â24 with n. 30, followed by Peters 1980: 28) that the form may well be a hyper-Lesbianism. Alternatively, ÏόμονÏÎµÏ could have an analogical o-vowel of the root if we assume that this became productive in the thematic aorist in Lesbian, starting from forms with a vocalized zero grade *CrÌ¥Câ > /CroC-/, such as á¼ÏÏοÏον, á¼Î´Ïομον. As for ÏόλαιÏι beside ÏαλάÏ, in spite of Francis (1974: 24 n. 32) and Peters (l.c.) it would be hazardous to conclude anything on the sole basis of this form, as its root has no clear Indo-European etymology. Moreover, accepting the above analysis of ÏόλαιÏι would imply that all other instances of the reflex *CRHVâ > /CaRV-/ in Lesbian were borrowed from Ionic, which seems highly unlikely.35
In sequences of the structure *CRHC, an anaptyctic vowel developed after the sonorant: *CRÉHC. This vowel was subsequently colored by the following laryngeal, yielding the well-known long-vocalic triple reflex *CRÄC, *CRÄC, *CRÅC: cf. âγνηÏÎ¿Ï âbornâ < *ǵnh1-tóâ, Att. ÏÏá¾±Ïá½¹Ï âsold, for saleâ < *prh2-tóâ, and ÏλÏÏá½¹Ï âbay, pale; greenâ < *ǵhlh3-róâ. This means that *CRÉHC merged with *CReHC. Again, all Greek dialects have the same reflexes,36 and in fact Greek probably shared this development with Phrygian, witness ÏλÏÏá½¹Ï beside Phryg. Î³Î»Î¿Ï ÏÎµÎ¿Ï âgoldenâ (cf. now Obrador Cursach 2019: 234).
A more difficult problem is the existence of disyllabic reflexes of *CRHC. This issue will not be treated in detail in this book, and it would merit an extensive treatment of its own.37 I will limit myself to a few basic observations concerning two questions: what evidence is there for a disyllabic reflex, and how was this reflex conditioned?
In my view, two of the most clear-cut pieces of evidence are the following.
-
ÏαÏá½±ÏÏÏ âto stirâ has a disyllabic reflex of zero grade *dhrÌ¥h2ghâ, while ÏÏá¾±Ïá½»Ï âroughâ < *dhrÌ¥h2gh-úâ shows the reflex with a long vowel. The alternative form of the present θÏá½±ÏÏÏ âto stirâ and the perfect Ïá½³ÏÏηÏε âis stirred upâ may reflect either *dhrÌ¥h2ghâ or full grade *dhreh2ghâ.
-
ÏαλάÏÏÏ âto soil, spatterâ is etymologically related to ÏλήÏÏÏ âto strike, blowâ (from a root *pleh2gâ, *plh2gâ), as argued in Van Beek 2013b.
These examples (and various other alleged cases) concern *h2, but what happened to the other two laryngeals? It has been suggested (e.g. Rix 1992: 73) that the disyllabic reflex was also threefold /eRe aRa oRo/, parallel to the long vowel reflexes, an alleged example being γένεÏÎ¹Ï âoriginâ < *ǵnÌ¥h1-tiâ. This development has been questioned by Peters (1980: 29). A form like γένεÏÎ¹Ï could ultimately reflect a full-grade form *ǵenh1â of the root (whether this was originally present in the paradigm or analogically introduced from elsewhere). The same holds for the aorist γένεÏο. Peters draws attention to ÏαÏá½³ÏÏη âquiverâ, which could reflect *bhrÌ¥h1-treh2â âmeans of carryingâ, closely resembling Ved. bharÃtraâ n. However, the Vedic form is a hapax, its meaning is not quite clear (possibly âfingerâ or âarmâ), and it could be a nonce formation based on carÃtraâ âlegâ (cf. EWAia s.v.). Thus, it is uncertain whether ÏαÏá½³ÏÏη may reflect *bhrÌ¥h1treh2â (with an extended root *bherh1â âto carryâ); whether *CRHC had a threefold disyllabic reflex remains an open issue.38
As for the factor conditioning the twofold reflex of *CRHC, it is widely believed that the disyllabic treatment was regular only in words with a secondarily retracted accentuation (e.g. Rix 1992: 73, Harðarson 1993). However, as Rix remarks, this requires that we assume a later accent shift in cases with an accented long vowel reflex such as κÏάαÏοÏ, κÏάαÏα < *krÄhatâ < *ḱrÌ¥h2snÌ¥-tâ âheadâ. This is conceivable, but not evident; excellent discussions of previous hypotheses and the problems involved can be found in Vine 1998: 66â69 and Probert 2006: 233â236. In my view (cf. Van Beek 2021b), the conditioning factor for the disyllabic reflex may have been the number of following consonants (i.e. disyllabic reflex in *CRHCC, long vowel reflex in *CRHCV), but this point will not be further pursued here.
1.2.2 * r̥ and * l̥ before a Glide
It is widely acknowledged that between a PGr. consonant and prevocalic glide, all Greek dialects regularly developed an a-vowel before liquids: *rÌ¥, lÌ¥ > αÏ, αλ | *C_ i̯V.39 The main pieces of evidence are the verbs in âαίÏÏ, which never turn up with o-vocalism (*âοίÏÏ) in any dialect, and feminine forms in âαιÏα such as ἰοÏέαιÏα âwho holds an arrow in her handâ (epithet of Artemis), probably reflecting PGr. *isu̯o-khesrÌ¥-i̯a.40 As with *CRHV, the development of the syllabic nasals is identical to that of the liquids: cf. á½Î½Î¿Î¼Î±á½·Î½Ï âto nameâ < *onomani̯e/oâ < *onomnÌ¥-i̯e/oâ. Thus, most scholars are inclined to posit a Common Greek development *CRi̯V > *CÉRi̯V > *CaRi̯V.
There are, however, some potential obstacles. First of all, there is the possible testimony of dialectal *âori̯â < *ârÌ¥i̯â in Aeol. á½Î½Î¿Î¹ÏÎ¿Ï âdreamâ (Sapph.). However, this form does not force us to reconstruct an old *onrÌ¥i̯oâ, as it might rather be a reshaping of PGr. *oneri̯oâ (as reflected in Class. á½Î½ÎµÎ¹ÏοÏ) under the influence of *á½Î½Î¿Ï, assuming that this was the regular Aeolic reflex of *onrÌ¥ (> Class. á½Î½Î±Ï).41
Secondly, it is hard to find instances of *âlÌ¥i̯â and *ârÌ¥i̯â that are demonstrably of Proto-Greek date. A possible example for *âlÌ¥i̯â is ἠλλομαι âto jumpâ, if this derives from *slÌ¥-i̯e/oâ and is to be directly compared with Lat. saliÅ âid.â. However, the a-vocalism of the Latin verb is difficult to obtain from a root *selâ, and may rather point to a root *sh2elâ (see EDL s.v.).42 As for *ârÌ¥i̯â, many verbs in âαίÏÏ can be analyzed as inner-Greek denominatives to stems in âαÏ, such as ÏεκμαίÏομαι âto conjectureâ beside Ïá½³ÎºÎ¼Î±Ï âsignâ. Likewise, a number of feminines in âαιÏα stand beside forms containing âαÏâ, such as ÏίμαιÏα âshe-goatâ beside ÏίμαÏÎ¿Ï âhe-goatâ. In such cases, the a-vocalism could have arisen first as a result of the word-final development *ârÌ¥ > âαÏ.
Notwithstanding these doubts, certain isolated formations strongly speak in favor of a Pan-Greek development to *rÌ¥ > Î±Ï before yod. The present stem formation of ÏαίÏÏ âto feel goodâ seems to have been inherited from earlier PIE *ǵhrÌ¥-i̯e/oâ, if we consider the cognate verbs Ved. háryati âid.â, Lat. horior âto encourageâ, U. heriiei âwishesâ.43 Another example of considerable antiquity could be μεγαίÏÏ âto begrudgeâ, cognate with Arm. mecarem âto hold in esteemâ, which probably continues the same pre-form and illustrates an intermediary stage of the semantic development leading from μέγα âbigâ to μεγαίÏÏ.
Another issue concerns the relation between the outcome *âari̯â (in the verbal formations just discussed) and the different syllabification found in forms like gen. pl. ÏÏιῶν âthreeâ or the feminine agent nouns in âÏÏια (already in Myc. âti-ri-ja, âti-ra2) < PGr. *âtria, *âtriÄâ < PIE *âtr-i(e)h2â. According to Ruijgh (1992: 78â¯ff.), the outcome in ÏÏιῶν and âÏÏια is regular, and the development to *âaRi̯â (in presents in âi̯e/oâ and motional feminines in âi̯a) occurred whenever *i̯ was analogically re-introduced, as a result of which the syllabic sonorant developed secondarily. However, the converse could also be defended: the syllabification reflected in ÏÏιῶν could be analogical after the dat. ÏÏιÏá½· or acc. *trins, and the feminine agent nouns in *âtr-ih2â also contain a morpheme boundary.44 In this connection, the form Ïá½¹Ïνια âladyâ (Myc. po-ti-ni-ja) is of prime importance. Since no base form with the stem *potnâ existed, Ïá½¹Ïνια (rather than Ïá½¹Ïαινα*) must display the regular reflex of *potni̯a (ultimately from PIE *potnih2). But even this is not the end of the story: Peters (1980) has argued that in the forms *âtri̯a and *potni̯a (reflected in âÏÏια and Ïá½¹Ïνια), the morpheme *âi̯a may have been restored, and that the sole example of a regular reflex of *âCri̯a would be á¼ÏÎ¿Ï Ïα âarable landâ (Myc. a-ro-u-ra) < *aro-u̯r-i̯a, with regular loss of yod in this environment.
Although these issues certainly merit a more detailed discussion, the exact scenario need not concern us here: the main point is that Greek furnishes no evidence for a prolonged retention of syllabic liquids before *i̯. Whenever *ârÌ¥i̯â arose in Proto-Greek, it seems to end up with an anaptyctic âaâ before the liquid in all Greek dialects. In this connection, the development of the syllabic nasals in the same environment is also relevant:45 in the inherited present stems Î²Î±á½·Î½Ï âto walkâ < *gwmÌ¥i̯e/oâ (cf. Lat. veniÅ âto comeâ) and μαίνομαι âto rageâ < *mnÌ¥i̯e/oâ (cf. Ved. mányate âthinksâ), we are clearly dealing with an early development *nÌ¥ (including original *mÌ¥) > *an before a glide, and analogical restoration is unlikely. There is every reason to think that cases like ÏαίÏÏ, with a-vocalism before the liquid, arose as part of the same development.
As for the outcome of the syllabic liquids in the environment PGr. *C_u̯V, it is difficult to cite a convincing example. The problem can be illustrated by an example containing a nasal: Î¼Î±Î½á½¹Ï âthin, sparseâ. This adjective has á¾± once in Empedocles (fr. 75.2 DK), but á¾° generally in Attic, and it therefore presupposes a pre-form *manu̯óâ. However, this *manu̯óâ probably does not reflect PGr. *mnÌ¥-u̯oâ, because the gloss μανύ (Hsch.) suggests that the form is due to the thematicization of an older u-stem, *mnH-uâ.46 Such a proto-form is corroborated by Arm. manr âsmallâ (gen. manu).47 The same type of formation may underlie Ion. κᾱλόÏ, Att. κᾰλόÏ, Boeot. καλÏοÏ, if my idea that the underlying root ended in a laryngeal (section 10.5.3) is correct.48 Finally, the neuter ÏᾶÏÎ¿Ï âclothâ (Hom.+), Myc. pa-we-a2 has been compared in previous scholarship with Lith. bùrva âcolor, colored garmentâ and bùrÄ âsailâ, but according to Fraenkel (LEW s.vv.), the former was probably borrowed from Polish barva, itself from MHG varwe (MoHG Farbe âcolorâ), and the latter is considered to be a loan from Finno-Ugric. Thus, since the etymology of *pharu̯os remains uncertain, it is unknown whether this word contains a reflex of *rÌ¥.49
1.2.3 Word-Initial * rÌ¥â and * lÌ¥â
A number of discussions of the development of the syllabic liquids in Greek distinguish a special Common Greek outcome á¼Ïâ in word-initial position.50 However, if the phonotactics of PIE did not allow a word-initial onset *râ, as seems likely, it is doubtful that syllabic *rÌ¥â existed in this position.51 Furthermore, it has become clear that many apparent cases of á¼Ïâ can or must be derived from a pre-form with PIE *h2râ.52 In word-initial *HLCâ, an epenthetic vowel developed in early Proto-Greek; in the ensuing *HÉLCâ the shwa was subsequently colored by the neighboring laryngeal.53 For instance, the following words may reflect full grade *h2erCâ or zero grade *h2rCâ:
-
á¼ÏÎºá½³Ï âto ward off, protect; be sufficientâ beside Hitt. ḫar(k)-zi âto hold; keepâ, Lat. arceÅ âto enclose; debar, keep away; protectâ;
-
á¼ÏÏá½»Ï âto arrange, prepareâ (cf. á¼ÏÏá½»ÏÎ ÏύνÏÎ±Î¾Î¹Ï âarrangement, orderingâ Hsch.) beside Lat. artus âjointâ, Ved. rÌ¥tú- m. âorder, fixed timeâ.54
Moreover, instances of á¼Ïâ may reflect *u̯rÌ¥Câ and, in forms with psilosis or Grassmannâs Law, *srÌ¥Câ. A possible case of *srÌ¥Câ is á¼ÏÏÏ âto be first; ruleâ. The etymology of this verb has been variously interpreted: *h2rÌ¥-ske/oâ with the root *h2erâ âto fitâ (Klingenschmitt 1974: 274 n. 1; accepted by Le Feuvre 2015: 506â507), or *rÌ¥gh-e/oâ to a root PIE *reghâ as reflected in MHG pret. rac âarose, protrudedâ and regen âto inciteâ (Tichy apud LIV2 s.v. *reghâ). In the latter analysis, it would be an instance of word-initial *rÌ¥â. In my view, á¼ÏÏÏ rather reflects PIE *sr̥ǵh-e/oâ or perhaps rather *sr̥ǵh-ske/oâ âto stand outâ, to the PIE root *serǵhâ reflected in Hitt. sarkuâ, sargau̯â adj. âpre-eminent, powerfulâ < *sr̥ǵh-(e)uâ, sarkiske/a-zi âto be eminentâ and Toch. B ṣärkâ âto surpassâ. This new proposal establishes á¼ÏÏÏ as an instance of the treatment of word-internal *rÌ¥ in Ionic-Attic. For further details, see section 9.6.2.
The only potential (though uncertain) piece of evidence for initial *rÌ¥â in Proto-Greek is á¼ÏÏην âmaleâ (Att. á¼ÏÏην, Thess. οÏÏεν, Arc. οÏεν, ÏοÏÏενÏεÏον), but there the place of the anaptyctic vowel may have been influenced by the variant á¼ÏÏην. This complicated example will receive further discussion in section 9.1.7.
There was no phonological constraint against word-initial *lÌ¥â, but there are no examples for its reflex in Greek. For instance, the root underlying á¼Î»Îºá½µ âmartial courageâ was *h2lḱâ with an initial laryngeal, as shown by the related á¼Î»á½³Î¾Ï âto ward offâ < *h2leḱ-sâ. In various words with initial λαâ, this sequence may reflect *lh2Câ, *slÌ¥â or *u̯lÌ¥â (see chapter 10).55 Ruijgh (1992: 86 n. 31) draws attention to λεÏÏá½¹Ï âdelicate, smallâ, a verbal adjective in âtoâ to λέÏÏ âto peel, scaleâ that is attested already in Myc. re-po-to. As one would expect a pre-form *lÌ¥ptóâ with zero grade root in this type of formation, roots of the structure *leCâ apparently generalized the full grade at an early date.
1.2.4 Word-Final * r̥ and * l̥
In word-final position, we only have evidence for *ârÌ¥; there are no clear examples of *âlÌ¥. I will postpone the discussion of word-final *ârÌ¥ to chapter 9, when we will have obtained a clearer picture of the word-internal developments. For now, let me briefly mention the two main issues. First, various scholars have posited an early, Pan-Greek change *ârÌ¥ > âÎ±Ï which took place prior to the vocalization of *rÌ¥ in word-internal position.56 Others assume that *ârÌ¥ would develop to either âÎ±Ï or âÎ¿Ï depending on the dialect, just as in word-internal position, and assume that only the place of the anaptyctic vowel (*Ér rather than *rÉ) was different in this position.57 We will be in a better position to overview the arguments once we have treated the evidence for the reflex of word-internal *rÌ¥ in Mycenaean and the alphabetic dialects (in chapters 2 and 3). Secondly, a special issue is the outcome seen in á½ÏόδÏα â(looking) sternlyâ < *upo-dr̥ḱ, which has been interpreted by Hoenigswald (1988) as evidence for *ârÌ¥ > âÏα. Barnes (2011), however, has ascribed the different reflex in á½ÏόδÏα to relative chronology, assuming that word-internal *rÌ¥ was vocalized before the loss of word-final stops. Again, we will be able to evaluate this piece of evidence more effectively once we have treated the word-internal evidence.
1.2.5 *r̥ and *l̥ before Nasals
Haug (2002: 54) has tentatively proposed that *rÌ¥ and *lÌ¥ developed a-vocalism in all Greek dialects in the environment *C_NV.58 His evidence consists of Lesbian and Thessalian ÏÏάλλα (Aeolic for ÏÏήλη âsteleâ) and the proper name Myc. wa-ni-ko. Although both examples mentioned by Haug have an outcome âaLâ, in his view the place of the anaptyctic vowel may have been either âÉLâ or âLÉâ.
Most handbooks do not treat this issue, which could be taken as an indication that they reject a special development for *rÌ¥n and *lÌ¥n. Indeed, the items ÏÏάλλα and wa-ni-ko do not prove the claim made by Haug. First of all, the reconstruction he proposes for ÏÏήλη is not certain. The handbooks compare it with OHG stollo âsupport, postâ (m. n-stem) and related Germanic forms; this would presuppose a pre-form form with *stlÌ¥-nâ.59 However, the alternative reconstruction of ÏÏήλη as *sth2-sleh2â by Risch (1974: 110; accepted also by Sihler 1995: 213) from the root *steh2â âstandâ cannot be excluded.60 Against the reconstruction *stlÌ¥-n-eh2â it can be objected that a verbal root *stelâ in the meaning âto standâ may be found in Germanic, but it does not exist in Greek (ÏÏá½³Î»Î»Ï means âto equipâ).
There are also phonological objections to a reconstruction *stlÌ¥nÄâ for ÏÏήλη, ÏÏάλλα. It is questionable whether a geminate resulting from *âlÌ¥nâ would have taken part in the first compensatory lengthening in Ionic-Attic (cf. section 10.5 on βάλλÏ, Îºá½±Î»Î»Î¿Ï and similar forms). Moreover, one expects *stlÌ¥nÄâ to develop o-vocalism in Aeolic dialects, even if there is no direct evidence for the outcome of *lÌ¥ in Lesbian (cf. sections 3.3.4 and 10.6). In view of these objections and of Rischâs alternative reconstruction, Aeolic ÏÏάλλα cannot be considered probative for Haugâs thesis.
As for Myc. wa-ni-ko, this is often interpreted as a diminutive reflecting *u̯rÌ¥n-iskoâ that would contain the stem of á¼Ïήν âlambâ. However, the root of á¼Ïήν must have been *u̯rh1â in view of ÏολύÏÏην ârich in lambsâ and Ved. úran- âlambâ, with a vocalic onset reflecting the root-final laryngeal. Therefore, the oblique stem á¼Ïνâ < *u̯arnâ must be analogical for earlier *u̯rÄnâ < *u̯rÌ¥h1-nâ, with *u̯arâ taken from the nominative *u̯arÄn < *u̯rÌ¥h1-Än. Thus, even if wa-ni-ko is to be connected with âlambâ (which is uncertain), it cannot be used to determine the regular reflex of *rÌ¥n.
In fact, there is ample further material for the development of *rÌ¥, lÌ¥ in the environment *C_nV. Most of the evidence is found in nasal present formations, where the vowel always appears before the liquid.61 These forms will be discussed in sections 9.4 (*ârÌ¥nâ) and 10.5 (*âlÌ¥nâ). Anticipating my conclusions, the evidence suggests that the vocalization of *ârÌ¥nâ and *âlÌ¥nâ took place in the individual dialect groups and cannot be ascribed to Proto-Greek.
1.2.6 Conclusions on Early Anaptyxis
-
Cases like á¼Ïήν âlambâ where a syllabic liquid allegedly stood before a vowel are in fact cases of the environment *CLÌ¥HV.
-
Before semivowels, *rÌ¥ was eliminated in Proto-Greek; it developed to âαÏâ before *i̯, at least when the sequence *ârÌ¥i̯â underwent morphological restoration. There is no secure evidence for *âlÌ¥i̯â, *âlÌ¥u̯â or *ârÌ¥u̯â.
-
The evidence formerly adduced for word-initial *rÌ¥â > Common Greek á¼Ïâ is obsolete in the light of the laryngeal theory.
-
The development of word-final *ârÌ¥ is still debated and will be discussed in section 9.4.
It is also important to distinguish chronological levels: the developments *CLÌ¥HV > *CaLV and *CrÌ¥i̯V > *Cari̯V took place at an early date, probably as early as Proto-Greek. In word-initial and word-final position, however, the dialect groups may have a diverging treatment. Nevertheless, as we shall see in section 9.5 there is strong evidence that word-final *ârÌ¥ developed earlier than word-internal *ârÌ¥â.
From now on, our main focus will be on the environments *CLÌ¥T (where *T is any occlusive or *s) and *CLÌ¥NV. Unless otherwise indicated, the debate about the âdouble reflexâ Î±Ï ~ Ïα in Ionic-Attic concerns these environments.
1.3 The oâ and u-Colored Reflexes of *rÌ¥ and *lÌ¥ in the Environment *C_T
In most dialects, the anaptyctic vowels in the reflexes of the syllabic liquids may appear with different qualities. For instance, in literary Lesbian we find both α (e.g. καÏδία, á½Î½Î±Ï) and ο (á¼Î¼Î²ÏοÏε, ÏÏÏá½¹ÏοÏ). Differences of opinion exist on the cause of these different reflexes, especially concerning the o-colored reflex in Mycenaean, Arcado-Cyprian and the Aeolic dialects. It has been debated whether the o-reflex in these dialects was the unconditioned outcome of PGr. *rÌ¥ and *lÌ¥, or whether it occurred only in some sort of labial environment. In section 1.3.1, I will give only a brief introduction to this problem; the evidence will be discussed in full detail in chapters 2 and 3.
Secondly, a few remarks will be devoted to the relatively marginal evidence for u-vocalism in Ionic-Attic (section 1.3.2). A third problem concerns the relation between the vocalization of the syllabic liquids and that of the syllabic nasals in those dialects which attest o-colored reflexes of both. In section 1.3.3, I will argue that these two developments must be viewed independently.
1.3.1 Which Dialects Have a Regular o-Colored Reflex?
As is well-known, o-colored reflexes of *rÌ¥ appear in Arcado-Cyprian and the Aeolic dialects, and Mycenaean also spells the outcome with signs of the o-series. The most important question is whether the o-colored reflexes are conditioned by their phonetic environment or, put differently, how serious the evidence for a-vocalism in these dialects really is. Since Morpurgo Davies (1968), it has been remarked time and again that the o-reflex frequently appears in a labial environment. Morpurgo Davies herself proposed a strict condition: only a preceding *u̯â would have conditioned the o-coloring in Arcado-Cyprian and Mycenaean, and the normal reflex of *rÌ¥ in these dialects would be ra or ar. However, anticipating the conclusions of chapters 2 and 3, I have not found a compelling reason to doubt an unconditioned o-colored reflex in these dialect groups, with the possible exception of Mycenaean, which may have preserved *rÌ¥.
Most scholars do not doubt that an unconditioned a-colored reflex is regular in Ionic-Attic and the West Greek dialects. A notable exception is Bader, but her suggestions have not been taken very seriously, probably because she did not try to establish a distribution between aâ and o-colored reflexes, and resigned to the conclusion that both reflexes may appear in any dialect without further conditioning (Bader 1969: 57â58).62
The potential instances of o-vocalism in Ionic-Attic will receive further discussion in chapter 9; for most of them alternative explanations are available. There is also one West Greek dialect that shows evidence for o-vocalism: as I will argue in chapter 3, in Cretan the development of *r̥ may have been conditioned by the preceding segment (labial versus non-labial).
1.3.2 The u-Colored Reflex
In various branches of Indo-European, the outcome of the syllabic liquids depended on surrounding consonants. In most cases where we find such a conditioned development, a preceding or following labial consonant colors the anaptyctic vowel to u. Thus, in Balto-Slavic the normal reflexes are ir, il, whereas convincing examples of ur, ul are found mostly after labiovelars.63 In Indo-Aryan, a similar conditioning determined the outcome of *LÌ¥ before laryngeals: contrast e.g. Ved. tirás âacrossâ < *trÌ¥h2nÌ¥s (cf. Lat. trÄns âid.â) with purás âbefore; in frontâ < *prÌ¥Hós (cf. Ïá½±ÏÎ¿Ï âbeforeâ).64 Phonetically, the anaptyctic vowel was rounded under influence of the preceding labialized stop.
It would not be surprising if similar effects were found in Greek. An extensive discussion of potential u-colored reflexes in Greek is Bernabé (1977: 275â283). However, most of his material concerns syllabic nasals and is therefore not directly relevant to our discussion, as these probably vocalized earlier than the syllabic liquids, and in a different way (cf. section 1.3.3 below). Words with u-vocalism that cannot be inherited must also be left out of consideration: well-known examples are Ïá½»ÏÎ³Î¿Ï âfortificationâ (cf. Goth. baurgs âcitadelâ) and Ïá½»Î¼Î²Î¿Ï âtumulusâ (cf. Ïá½±ÏÎ¿Ï âburialâ), which are usually considered to be borrowings from a different, unattested Indo-European language into Greek.
Other reconstructions mentioned by Bernabé do not strike me as particularly convincing, for instance:
-
κύÏÏÎ¿Ï âfishing netâ (Sapph.+) has been reconstructed as *krÌ¥toâ and compared to Proto-Germanic *hurdiâ âwickerwork (door)â (Goth. haurds âlatticed doorâ, OHG hurt âhurdle, grate, railingâ and other forms, cf. EDPG s.v.). However, the type of referents of these words renders any etymology open to doubt, and the Germanic words are more likely to be related to Lat. crÄtis âhurdleâ < *krh2-tiâ, with a root that cannot account for the Greek form.
-
Ïύλλον âleafâ beside Lat. folium âid.â. The two must not be reconstructed as *bhlÌ¥i̯oâ (as per Bernabé 1977: 283) but may rather reflect *bholi̯oâ, with a raising *o > Ï before *âli̯â as proposed by Vine (1999: 564â569).
-
ÏÎºá½»Î»Î»Ï âto tear apart, snatchâ (A.+) and ÏÎºá½±Î»Î»Ï âto hoe; stir upâ (Hdt.+) are supposed to be vocalizations of *sklÌ¥-i̯e/oâ or *skÉl-i̯e/oâ with a âreduced gradeâ by the etymological dictionaries. However, upon this account (accepted by Bernabé 1977: 277) it would be difficult to account for the two divergent reflexes. If there is indeed an etymological connection,65 one might follow Vine (1999: 566) in reconstructing a pre-form with o-grade for ÏκύλλÏ, with the same raising as in Ïύλλον.
-
The comparison between ÏÏῦÏα âhammerâ and ÏÏαá¿Ïα âballâ (Bernabé 1977: 283) does not seem cogent to me in view of the semantic divergence; there are no ascertained cognates outside of Greek.
Nevertheless, even if such cases are left aside, some interesting candidates to show an u-colored outcome remain. I will now first discuss two promising cases: Î»á½»ÎºÎ¿Ï âwolfâ and the adjective ÎºÏ ÏÏá½¹Ï âhumpedâ.66
Usually, Î»á½»ÎºÎ¿Ï is considered an instance of metathesis from PIE *u̯lÌ¥kwoâ. However, when *u̯lÌ¥kwo- came to be realized as *u̯lÉkwoâ, it would be natural that the anaptyctic shwa was rounded due to the presence of labialized consonants on both sides, after which *u̯lukwoâ yielded λύκοÏ. The chronology is unproblematic: the rounding may have taken place before [lÉ] developed into λα (the regular reflex, cf. chapter 10). The rounding of an anaptyctic vowel and subsequent delabialization is paralleled by Î³Ï Î½á½µ âwoman; wifeâ, which no doubt reflects *gwÉnÄ. Moreover, compare the reflex of a syllabic nasal in Ïύξ adv. âwith the fistâ < *pnÌ¥kw-s (van Brock 1972), with the root of âfiveâ and related to OE fyst âfistâ, OCS pÄstÑ âid.â < *pnÌ¥kwstiâ.67 Not all anaptyctic vowels in the prehistory of Greek were rounded by neighboring labiovelars; the development is not found in early instances of anaptyxis preceding the loss of laryngeals (e.g. βαÏá½»Ï < *gwÉruâ < *gwrH-uâ). This means that the anaptyctic vowel in *gwÉnÄ developed after *gwÉruâ had become *gwaruâ.68 A late date of *gwnÄ > *gwÉnÄ is corroborated by Boeot. βανά, where the anaptyctic shwa apparently developed after the elimination of the labiovelars. At any rate, explaining Î»á½»ÎºÎ¿Ï as a regular vocalization of *u̯lÌ¥kwoâ would be an attractive alternative over assuming an irregular metathesis. Chronologically, this would place the anaptyxis after *lÌ¥ before the disappearance of labiovelars and before the reduction *u̯lâ > λâ, i.e. in or before the Mycenaean period.
A second case is the adjective ÎºÏ ÏÏá½¹Ï âbulging (of a wave); humped (of shoulders), hunchbackedâ (Il.+), later âconvexâ. Its root has been compared to that of Lat. curvus âcurved, convexâ, but as De Vaan (EDL s.v. curvus) remarks, *kurâ is not an allowed PIE root structure, and there is no PIE root *kwer- meaning âturn; roundâ vel sim. However, whether or not the etymological connection with Lat. curvus is correct, root structure constraints do suggest a reconstruction *kwrÌ¥-tóâ for ÎºÏ ÏÏá½¹Ï.69 In fact, the root *kwerâ âto cut off, amputate, mutilateâ is an excellent candidate, as various of its derivatives denote corporeal defects, e.g. Ved. kará¹Ã¡- âcrop-earedâ, CS krÑnÑ âmutilated (with ears slit or cropped)â, Sln. krÌn âmaimed, mutilatedâ (from Proto-Slavic *kÑrnÑ âmaimedâ).70 The meaning âhumped; hunchbackedâ of ÎºÏ ÏÏá½¹Ï may easily have developed from âtruncated, bluntâ. Furthermore, it is attractive to compare ÎºÏ ÏÏá½¹Ï directly with Lat. curtus âmutilatedâ, equally from *kwrÌ¥-tóâ, and with Lith. kurÌÄias âdeafâ, kurÌtas âid.â.71 This provides ÎºÏ ÏÏá½¹Ï with a semantically attractive etymology which also explains its u-vocalism.
In addition to these two forms, three more potential (but rather complex) examples must be discussed:
-
Ïá½»Ïξ âmeatâ, which is mentioned as the Aeolic and Doric form of Ïá½±Ïξ âid.â (< PGr. *tu̯rÌ¥kâ) in the Etymologicum Magnum,72 and as Aeolic in Ïá½»ÏκεÏιΠÏαÏξίν. Îἰολεá¿Ï (Hsch.);
-
Ïá½»ÏÏ âto draw, dragâ (Ion.-Att.), which is surely related to ÏαίÏÏ âto sweepâ (S.+, also epigraphically in Cretan) < PGr. *tu̯rÌ¥-i̯e/oâ;
-
PN Î¤Ï ÏÏαá¿Î¿Ï, which is often supposed to mean âborn on the fourth dayâ, and therefore thought to derive from *ÏÏ ÏÏá½µ âfourth dayâ, which would continue a relic form of the ordinal PGr. *kwtu̯rÌ¥tóâ âfourthâ.
Concerning Î¤Ï ÏÏαá¿Î¿Ï, we must take into account that the first part of ÏÏá½±Ïεζα âtableâ is usually thought to derive from *kwtu̯rÌ¥â as well.
In all these examples we are faced with the problem of accounting for a double reflex of *u̯rÌ¥: in part of the cases *rÌ¥ appears to have undergone its normal vocalization, but in other cases there seems to have been a re-vocalization of *u̯rÌ¥ to *ur. If that is indeed what happened, we should be able to indicate a phonetic factor that caused this re-vocalization. On the other hand, we must reckon with the possibility that Ï Ï reflects not *u̯rÌ¥, but something else. What follows now is an attempt to make sense of the data.
Starting with Ïá½±Ïξ beside Ïá½»Ïξ, the latter form may reflect an o-grade *tu̯órkâ, with raising due to Vineâs version of Cowgillâs Law (Vine 1999: 570â572, elaborating a suggestion by Schindler 1972: 34), whereas Ïá½±Ïξ would contain the normal zero grade reflex.73 This possibility suffices to eliminate Ïá½»Ïξ from the compelling evidence for re-vocalization of *u̯rÌ¥.
As for Ïá½»ÏÏ beside ÏαίÏÏ, the evidence for a PIE root *tu̯er- âto sweep, rushâ includes at least the following forms:
-
pres. *tu̯er-e/oâ > PGmc. *þweranâ (strong verb) > OE þweran âto twirl, stirâ, OHG dweran âto stir upâ, and also Ved. tvárate âto rushâ.
-
pres. *tu̯rÌ¥-i̯e/oâ reflected not only in ÏαίÏÏ, but also in ON þyrja âto sweep, rushâ (Kroonen, EDPG s.v. *þurjanâ).
Until the treatment by Vine (1999: 570), the twofold outcome of *tu̯rÌ¥-i̯e/o- in Greek (ÏαίÏÏ, Ïá½»ÏÏ) was usually left unexplained. As Vine remarks, however, the two reflexes cannot be ascribed to dialectal differences, as both verbs are attested in Ionic-Attic. If ÏαίÏÏ reflects the regular treatment of *tu̯rÌ¥-i̯e/oâ, as seems likely (with an early, Common Greek vocalization *ârÌ¥â > âarâ before yod), what does Ïá½»ÏÏ reflect? I will discuss two options.
First, one could analyze Ïá½»ÏÏ as a secondary denominative based on nominal forms like *ÏÏ ÏÏá½¹Ï âstirred, in sweeping motionâ (cf. κολοÏÏ ÏÏá½¹Ï âsweeping motion, tumultâ Il.+, although the element κολοâ remains enigmatic; cf. DELG s.v.). This *ÏÏ ÏÏá½¹Ï could be the regular outcome of *tu̯rÌ¥-tóâ if we assume (i) that *tu̯rÌ¥â > *tsu̯rÌ¥â would have yielded ÏÏ Ïâ before an occlusive, whereas (ii) before yod, *tu̯rÌ¥â was vocalized as *tu̯Érâ > *tu̯arâ at an earlier date. In other words, the development would be conditioned by the type of consonant that followed. However, the form Ïá½±Ïξ is left unaccounted for in this scenario, which predicts that *tu̯rÌ¥ks would develop into Ïá½»Ïξ even in Ionic-Attic. One could surmise that the re-vocalization took place only in pretonic position, hence *tsu̯rÌ¥tóâ > *tsurtóâ whereas *tu̯rÌ¥Ìks was retained and later yielded Ïá½±Ïξ. However, for want of further examples, this is mere speculation, and one might also doubt the reconstruction *tu̯rÌ¥-tóâ underlying âÏÏ ÏÏá½¹Ï altogether.
A second possibility to account for Ïá½»ÏÏ emerges when we consider that Ïá½»Ïξ could reflect an o-grade *tu̯órkâ (cf. above) whereas Ïá½±Ïξ would contain the normal zero grade reflex. Vine leaves open the morphological motivation of the o-grade in Ïá½»ÏÏ, but I wonder whether âÏÏ ÏÏá½¹Ï (in κολοÏÏ ÏÏá½¹Ï âsweeping motionâ) may reflect an o-grade action noun *tu̯ór-toâ of the type Ïá½¹ÏÏοÏ.74 The assibilated form *tsu̯ór-toâ may have developed into *tsu̯urtoâ by Cowgillâs Law, and then simplified to *tsurtoâ. This presupposes that the anaptyctic vowel in *tsu̯Ér-i̯e/oâ (> ÏαίÏÏ) had already developed into a before Cowgillâs Law affected the o-grade (otherwise, it would be difficult to understand why a shwa was not rounded in the same environment), but this is an unproblematic assumption. We would then arrive at the following relative chronology:
| (1) |
*ârÌ¥i̯â > *âari̯â and assibilation *tu̯ > *tsu̯, in either order (yielding *tsu̯ari̯e/oâ beside *tsu̯órtoâ, *tsu̯rÌ¥kâ beside *tsu̯órkâ); |
| (2) |
Cowgillâs Law operates in the context *tu̯or (perhaps more generally *Tu̯oR, cf. Vine 1999), probably followed soon by a simplification *tsu̯â > *tsâ before u (yielding *tsúrtoâ, *tsúrkâ); |
| (3a) |
Initial *tsu̯â > *tsâ elsewhere, yielding *tsari̯e/oâ (> ÏαίÏÏ), *tsrÌ¥kâ (> Ïá½±Ïξ); |
| (3b) |
Creation of denominative *tsuri̯e/o- (> Ïá½»ÏÏ). |
Thus, it cannot be excluded that ÏÏ Ïâ in the forms Ïá½»Ïξ and Ïá½»ÏÏ reflects pre-forms with an o-grade (*tu̯orkâ, *tu̯orâ).
Concerning Î¤Ï ÏÏαá¿Î¿Ï, assuming that the derivation of this name from âfourthâ is correct, the expected vocalization of the ordinal would be *kwtu̯rÌ¥tóâ. In Proto-Greek, this would undergo simplification of the onset to yield *tu̯rÌ¥tóâ, and the absence of assibilation in this form could be accounted for with the assumption that *kwtu̯â was still intact when *tu̯â > *tsu̯â. Could it be that this *tu̯rÌ¥tóâ âfourthâ was re-vocalized as *turtó-? From a purely phonetic viewpoint this is conceivable, but the idea seems contradicted at least by Ïá½±Ïξ reflecting *tu̯rÌ¥kâ. Moreover, it must be taken into account that the first part of the word for âtableâ, attested as ÏÏá½±Ïεζα and Myc. to-pe-za, is usually reconstructed as *kwtu̯rÌ¥â as well. In that word, however, there is no trace of the putative *u̯.75
To be sure, it would be possible to resolve these issues. The reconstruction of âtableâ as referring to a four-legged object is not certain, as we will see in chapter 2. For the word for âmeatâ, one might assume that an earlier *turkâ was reshaped as *tu̯rÌ¥kâ under the influence of a full grade *tu̯órkâ. This leaves us with Î¤Ï ÏÏαá¿Î¿Ï, but it would not be prudent to base far-reaching conclusions on the interpretation of a personal name.
To conclude this section, ÎºÏ ÏÏá½¹Ï < *kwrÌ¥tóâ and Î»á½»ÎºÎ¿Ï < *u̯lÌ¥kwoâ may well display rounding of an anaptyctic shwa next to a labiovelar. One of these examples concerns *rÌ¥, the other *lÌ¥, so there is not much evidence to go by. The phonetic environment in *u̯lÌ¥kwoâ is highly specific: in this word, the syllabic liquid was flanked by two labialized sounds. Nevertheless, as far as I have seen there is no counterevidence against a development *KwrÌ¥â > *Kurâ (*Kw = any labiovelar), as the vocalism of βÏÎ±Î´á½»Ï âslowâ from *gwrÌ¥d-úâ may be analogical (see chapter 4).
1.3.3 The o-Colored Reflex of the Syllabic Nasals
The Greek vocalization of the syllabic liquids is often compared with that of the syllabic nasals.76 The rationale behind this comparison is that all syllabic sonorants may be reflected with either aâ or o-vocalism, and that the Greek dialects which generally have o-colored reflexes of *rÌ¥ also have instances of o-vocalism from *nÌ¥ or *mÌ¥.
There are, however, also some important differences between the regular development of *r̥ and that of the syllabic nasals. The following brief discussion will not solve all problems concerning the syllabic nasal reflexes; the goal is merely to argue that the changes affecting the syllabic liquids were chronologically later, and therefore best considered independently.
First of all, we must note that the similarities between the two changes are only superficial. The unconditioned regular reflex of *rÌ¥ is âÏοâ in the Aeolic dialects, âοÏâ in Arcadian (see chapter 3), and either rÌ¥ or âorâ in Mycenaean (chapter 2). However, there is no dialect which has âoâ as the unconditioned reflex of the syllabic nasals: the normal reflex of *nÌ¥, *mÌ¥ in all Greek dialects, including Mycenaean, is a.77 This fact by itself suffices to show that we are dealing with two distinct developments. Furthermore, the phonetics underlying the two developments are different. The nasal feature completely disappeared when *nÌ¥, *mÌ¥ were vocalized, probably through an intermediary stage [ÉÌ]. On the other hand, *rÌ¥ and *lÌ¥ were vocalized due to the phonemicization of an anaptyctic vowel; in this process the liquids were retained as independent segments.78
The main problem is to explain the conditioning of the reflex *nÌ¥, *mÌ¥ > o. Many scholars accept the thesis, first formulated for Mycenaean by Risch (1958: 160 n. 40) and taken up by Morpurgo Davies (1960), that the o-colored reflex is due to a neighboring labial consonant.79 The strongest examples for this development are a-no-wo-to /anohwoto-/ âwithout handlesâ < *an-ou̯snÌ¥-toâ (cf. Hom. οá½Î±Ïα âearsâ) and e-ne-woË ânine-â < PGr. *en(n)eu̯nÌ¥ (Class. á¼Î½Î½á½³Î±).80 Conditioning by a neighboring labial consonant would also explain why Myc. has a-mo âwheelâ (also nom. pl. a-mo-ta, dat. pl. a-mo-si) corresponding to Hom. á¼ Ïμα, á¼ ÏμαÏα âchariotâ, and pe-mo âseedâ (if this represents /spermo/ rather than /spermÅn/) beside alphabetic ÏÏá½³Ïμα âid.â.81 It could also explain the Homeric forms á½ÏαÏÏÎ¿Ï âof the same paternal ancestryâ < *smÌ¥-ph2tr-oâ and οἰεÏá½³Î±Ï âborn in the same yearâ < *smÌ¥-u̯etesâ, assuming that such forms are of Mycenaean origin.82 However, these forms with copulative á½â cannot carry too much weight: Ruijgh (1961: 201) explains them by an analogy that would have taken place in a psilotic dialect like Lesbian, where the preconsonantal variants á¼â < *smÌ¥â and á¼â < *nÌ¥â had merged. This would have motivated the analogical creation of á½â beside prevocalic á½Î¼â (< *som(o)â, á½Î¼Î¿â before consonants) after the model of á¼â beside prevocalic á¼Î½â.
There is, however, a severe problem with the idea of labials as a conditioning factor. A variation a ~ o is found in a small number of Mycenaean neuters. The forms with âaâ are:
-
pe-ma /sperma/, found also at Pylos, but only in one scribal hand and beside the much more frequent pe-mo;
-
AREPA /aleiphar/ âunguentâ, dat. a-re-pa-te /aleiphatei/; a-re-pa-zo-o /aleipha-ddjoho-/ âunguent boilerâ but also a-re-po-zo-o /aleipho-ddjoho-/;
-
ins. sg. e-ka-ma-te /hekhmatÄ/ and pl. e-ka-ma-pi /hekhma(p)phi/ < *hekh-mnÌ¥tâ âwith support(s)â (part of a table), to be compared with Hom. á¼Ïμα âsupport, propâ (of a ship or a wall).
These forms constitute a well-known crux of Mycenaean studies.83 I will not attempt to review all previous solutions, but instead discuss Ruijghâs solution in more detail, as it is well-known and directly related to his views on the syllabic liquids.84
Since Mycenaean also shows neuters in /-ma/,85 Ruijgh argued that the syllabic nasals had a-colored reflexes in all Greek dialects. In his view, the above-mentioned o-colored reflexes in Mycenaean originated in heteroclitic stems with nom.-acc. *ârÌ¥, gen. *ânÌ¥tos. Word-final *ârÌ¥ would have regularly yielded âor in âAchaeanâ and Aeolic dialects (i.e. the same vowel color as in word-internal position, but a different slot), while *ânÌ¥tos developed into *âatos, as elsewhere.86 He adduces the Homeric words ἦÏÎ¿Ï âheartâ and á¼Î¿Ï âswordâ, which in his view are âAchaeanâ elements of Epic Greek, as evidence for this development. Next, the heteroclitic paradigm could be leveled in two different directions. In literary Lesbian, a paradigm with âαÏ, âαÏÎ¿Ï was the result, and Ruijgh sees the same levelling reflected in Mycenaean AREPA, a-re-pa-te, a-re-pa-zo-o.87 In addition, he supposes that these heteroclitic stems had âdoubletsâ in *âor, *âotos in Mycenaean, which arose by leveling in the opposite direction. This assumption allows him to explain the o-vocalism in words like a-mo and pe-mo: the âdouble flexionâ (oblique forms in âatâ beside âotâ) would have secondarily spread to non-heteroclitic neuters in *âmnÌ¥(t), and finally even to the nom.-acc. sg. of such forms.88 Thus, alleged traces of such âdoubletsâ are pe-mo beside pe-ma and a-re-po-zo-o beside a-re-pa-zo-o, while e-ka-ma would have retained the original a-vocalism and a-mo generalized the âdoubletâ with âotâ.
This construction fails to convince for several reasons. First, as remarked by Cowgill, it is unlikely that the heteroclitic stems (a relic type) influenced a highly productive type like the neuters in *âmnÌ¥(t), âespecially when that influence consists in the creation of new doublets, rather than the favoring of one or another inherited form or the leveling of some anomalyâ (Cowgill 1966: 90). Secondly, it is implausible that two suffix variants âatâ and âotâ served as the productive marker of a morphological category in one single dialect. Thirdly, Ruijghâs scenario does not explain the distribution between âmo and âma, and it is particularly problematic that only a marginal lexical item like e-ka-ma would have resisted analogical reshaping. Finally, there is no unambiguous proof that heteroclitic neuters in *ârÌ¥ ever had a nom.-acc. sg. in /-or/ in Mycenaean: the evidence rather points to /-ar/, and the dialectal origin and reconstruction of Homeric ἦÏÎ¿Ï and á¼Î¿Ï remains uncertain (see section 9.5). Ruijghâs scenario therefore cannot be correct.
In sum, it still seems more likely that the o-colored Mycenaean reflex of syllabic nasals was conditioned at least by surrounding labials. This would explain several isolated forms, as well as the pervasive o-vocalism of Myc. a-mo < PGr. *ar-mnÌ¥ and the fact that pe-mo is the normal form at Pylos (with the exception of one single hand). However, it must be admitted that e-ka-ma and the variation pe-ma ~ pe-mo continue to present problems, and that there may have been other conditioning environments for the o-colored reflex.89 Finally, it must be noted that *mÌ¥ normally does not have a different reflex from *nÌ¥ in Mycenaean (nor in Aeolic dialects): for the development *mÌ¥ > a, cf. Myc. a2-te-ro /hatero-/ ânext [year]â < *smÌ¥teroâ âthe other [year]â.
In the remainder of this work, the syllabic nasals will only play a marginal role. In my explanation of numeral forms like Myc. qe-to-roâ, Thess. ÏεÏÏοâ, Class. ÏεÏÏαâ (section 2.7), I will argue that these dialectal differences can be explained by analogy with the corresponding differences in ânineâ and âtenâ.90 Furthermore, in chapter 9 it will be argued that certain instances of âÏαâ reflect pre-forms of the shape *CrnÌ¥C, e.g. γÏá½±Ï âto eat, digestâ.
1.4 Previous Accounts of âαÏâ versus âÏαâ in Ionic-Attic
Let us now turn to the central issue of this study, the double reflex âαÏâ versus âÏαâ of Proto-Greek *rÌ¥ in Attic and Ionic dialects. The claim defended in this book is that only âαÏâ was the regular reflex of Proto-Greek *rÌ¥. From the perspective of the Greek evidence usually marshalled, this is an unexpected result. However, if we compare the reflexes of *rÌ¥ in other Indo-European languages, we find that the anaptyxis normally occurs in front of the vowel: cf. PGmc. *ur, ul, Arm. ar, PAnat. ar, PToch. är, Proto-Balto-Slavic *ir/ur, *il/ul. The only branch of Indo-European showing regular anaptyxis after the liquid is Celtic: the Proto-Celtic reflexes are *ri, li (though only before stops and m: in other contexts the reflex is *ar).91 Within Greek, as we will see in chapters 2 and 3, only the Aeolic dialects provide clear evidence for an anaptyctic vowel developing after the liquid. Seen in this light, the claim that *rÌ¥ regularly yielded âαÏâ already appears to be much less outlandish.
An instructive treatment of the evidence in Ionic-Attic is KuryÅowicz (1968: 247), who cites the following evidence in favor of a regular development to âÏαâ:
Dass âÏαâ lautgesetzlich ist, beweisen Gegensätze wie δέÏκομαι : á¼Î´Ïακον; Ïá½³ÏÎ¸Ï : á¼ÏÏαθον; νημεÏÏá½µÏ aber á¼Î¼Î²ÏαÏαιΠεἵμαÏÏαι und á¼Î¼Î²ÏαμένηΠεἱμαÏμένη (Hesych);92 Ïá½³ÏÏÏ : ÏÏαÏείομεν; ÏÏείÏÏ : âÏÏÏαÏá½¹Ï; δέÏÏ : δÏαÏá½¹Ï; Ïá½³ÏÏομαι : ÏÏαÏιά; θέÏÏÎ¿Ï (äol.) : θÏαÏá½»Ï; Ïá½³ÏÏαÏÎµÏ (für *Ïá½³ÏÏοÏεÏ) : ÏÏá½±Ïεζα. Vgl. ferner isolierte Beispiele wie βÏÎ±Î´á½»Ï : lat. gurdus; κÏá½±Î½Î¿Ï : lat. cornus; ÏÏá½±Ïον : lat. porrum.
Although not all these examples are equally compelling, this is certainly an impressive list. One fact deserves special attention: whenever CraTâ appears as the zero grade reflex of a root of the structure CerTâ, it usually cannot be explained by analogical mechanisms.
The apparent impossibility to give a different explanation for âÏαâ in such paradigms as δέÏκομαι : á¼Î´Ïακον and Ïá½³ÏÎ¸Ï : á¼ÏÏαθον has always strengthened the conviction that âÏαâ is the regular reflex of *rÌ¥ in Ionic-Attic. This is, however, not the end of the story. As was recognized long ago, there are also cases of âαÏâ < *rÌ¥ in roots of the structure CreTâ. Osthoff (1879: 144â145) and Güntert (1916: 72) drew attention to κάÏÏα âveryâ beside κÏαÏá½»Ï (κÏá½±ÏοÏ, κÏαÏá½³Ï, κÏαÏαιόÏ), as well as ÏαÏÏá½»Ï, ÏαÏÏειαί âdense, frequentâ, Ïá½±ÏÏÎ¿Ï âthicketâ beside ÏÏá½³ÏÏ. Güntert eventually dismissed κάÏÏα in view of the possibility that Goth. hardus âhardâ is etymologically related, and waved away ÏαÏÏá½»Ï and Ïá½±ÏÏÎ¿Ï with the claim that they are artificial epic creations.93 Neither of these claims can be substantiated: κάÏÏα clearly belongs to the root κÏεÏâ, with a different full grade slot, and if ÏαÏÏá½»Ï would have been preferred over *ÏÏαÏá½»Ï for metrical reasons, it remains unclear why a similar reshaping did not take place in other Homeric adjectives like βÏαδύÏ, θÏαÏá½»Ï, κÏαÏá½»Ï.
KuryÅowicz dealt with κάÏÏα and ÏαÏÏá½»Ï by assuming that the fluctuation between zero grades CRaTâ and CaRTâ in roots of the structure CeRTâ induced a hesitation about the correct zero grade of roots of the structure CReTâ.94 It remains unclear, however, why hesitation about the correct zero grade would occur at all in roots with an otherwise unambiguous full grade slot. KuryÅowicz does not explain why this âmorphologically conditionedâ âαÏâ is found precisely in καÏÏεÏá½¹Ï, κάÏÏα and ÏαÏÏá½»Ï and not in other forms, nor why there are no by-forms xκÏá½±Ïα and xÏÏαÏá½»Ï. It is difficult, then, to dismiss κάÏÏα, καÏÏεÏá½¹Ï, and ÏαÏÏá½»Ï so easily as Güntert and KuryÅowicz did.
Another important problem concerns the existence of doublets of the type καÏÏεÏá½¹Ï ~ κÏαÏεÏá½¹Ï. In addition to words with the root καÏÏ- ~ κÏαÏâ, KuryÅowicz (1968: 247) mentions the following pairs of forms:
μάÏÏÏÏ und βÏá½±Ïαι; βάÏναμαι (neben μάÏναμαι) setzt *βÏάναμαι voraus; ÏαÏδεá¿Î½ : ÏÏαδεá¿Î½; καÏδία : κÏαδία; á¼Î´Î±Ïθον : á¼Î´Ïαθον (poet.); (â¦) ÏαÏÏá¿Î½Î±Î¹ (â¦), aber ÏÏαÏείομεν.
These examples are either true doublets (attested with both âαÏâ and âÏαâ), or could be taken to suggest the earlier existence of a doublet. Güntert (1916: 71â72) adduced several further examples, but most of them do not survive closer scrutiny;95 KuryÅowicz rightly restricted himself to a group of more central examples.
In all the cases listed above, KuryÅowicz views the form with âÏαâ as presenting the older reflex. Starting from the idea that âÏαâ is the regular reflex of *rÌ¥, most previous accounts use one or more additional hypotheses in order to explain the occurrence of forms with âαÏâ that cannot be due to analogical restoration.96 The following are the most notable proposals:
-
There was originally free variation between âÏαâ and âαÏâ (or, before the phonologization of shwa, between [rÉ] and [Ér]). Eventually, one of these variants was generalized in each lexeme, but in some cases older variants were preserved, especially in poetry.97
-
An accent-conditioned development, according to which only (secondarily) accented *rÌ¥Ì would yield âá½±Ïâ.98
-
Liquid metathesis of âÏαâ and âÏοâ yielded âαÏâ and âοÏâ, respectively.99
-
/CaRT-/ replacing /CRT-/ is a secondary (analogical or morphologically conditioned) ablaut variant of /CeRT-/ that arose before the vocalization of *RÌ¥.100
-
-αÏâ is the regular reflex only after heavy onset clusters, âÏαâ elsewhere.101
I will now discuss these proposals and the problems with them one by one.
1.4.1 Free Variation between âÏαâ and âαÏâ at an Early Stage
In his Grammaire homérique, Chantraine observed that doublets of the type καÏÏεÏá½¹Ï ~ κÏαÏεÏá½¹Ï are mainly found in Homer. He mentions the examples shown in Table 2 on the opposite page. Chantraine (1958: 23) proposes that these pairs originated as follows:102
Dans le développement des sonantes rÌ¥ et lÌ¥ lââ¯Ã©lément consonantique sââ¯est maintenu et lââ¯Î± figure soit avant soit après la consonne: on observe un flottement entre Ïα et αÏ. Les aèdes ont naturellement choisi la forme la plus favorable à lââ¯hexamètre dactylique.
Phrased in this way, Chantraine seems to accept both âαÏâ and âÏαâ as regular outcomes of *rÌ¥. He does so in order to explain the choices apparently available to epic singers. This would not explain, however, why the variation occurs only in these specific words, and thus it would amount to a resignation to the problem.103 On a more charitable reading, Chantraine may be taken to assume that at some point, before the vocalization was phonologized, forms with [rÉ] and with [Ér] were in competition. Only Epic Greek would preserve traces of the hesitation between these two competing realizations, and only in a small number of cases both variants were retained, because of their metrical utility.
This idea deserves attention because it would explain why variation between Ïα and Î±Ï in the same lexeme is practically limited to Epic Greek.104 It is, however, not without problems. First of all, it entails that variation between forms like καÏÏεÏá½¹Ï and κÏαÏεÏá½¹Ï existed already before the vocalization of *rÌ¥ (i.e. before the phonologization of one of the supposed variants [rÉ] and [Ér]). Since this stage is normally dated to before the Mycenaean period, this scenario would require a very long time depth for the epic tradition. Secondly, one would like to see other compelling reasons for assuming a true hesitation between the phonetic realizations [Ér] and [rÉ] in spoken prehistoric Greek. Thirdly, admitting that the variation between Ïα and Î±Ï is an artificial phenomenon does not explain the presence of forms with the reflex Î±Ï (like καÏÏεÏá½¹Ï) in the Attic vernacular.
Table 2
Doublet forms in which Î±Ï alternates with Ïα in Homer
|
Ionic-Attic |
Homeric |
|---|---|
|
Att. καÏδία, Ion. καÏδίη âheartâ |
κÏαδίη ~ καÏδίη |
|
θάÏÏοÏ, Att. θάÏÏÎ¿Ï âcourageâ (but also Att. θÏá½±ÏÎ¿Ï âaudacityâ) |
θάÏÏÎ¿Ï ~ θÏá½±ÏÎ¿Ï |
|
Ion.-Att. καÏÏεÏá½¹Ï âstrongâ |
κÏαÏεÏá½¹Ï ~ καÏÏεÏá½¹Ï |
|
Ion.-Att. Ïá½³ÏαÏÏÎ¿Ï âfourthâ |
Ïá½³ÏαÏÏÎ¿Ï ~ Ïá½³ÏÏαÏÎ¿Ï |
|
Att. βÏÎ±Î´á½»Ï âslowâ |
βÏαδύÏ, superl. βάÏδιÏÏοÏ. |
If the creation or retention of pairs like καÏÏεÏá½¹Ï ~ κÏαÏεÏá½¹Ï were due to metrical convenience only, it would be difficult to understand why such variation was exploited only on a limited scale. In the course of this book, we will repeatedly focus on the distribution of forms with âαÏâ and âÏαâ, especially among forms containing the root κÏαÏâ ~ καÏÏâ, and we will encounter various salient distributions. Such details remain unexplained if we assume that the poets could choose whichever variant they liked. For this reason, explanations along the lines of Chantraine are unlikely to be correct.
1.4.2 Accent-Conditioned Development
Schwyzer, in his Griechische Grammatik, recommends the following explanation (1939: 342):
Für καÏδία (aber hom. κÏαδίη, vgl. air. cride), θαÏÏεá¿Î½ (neben θÏαÏá½»Ï), δαÏÏá½¹Ï (neben δÏαÏá½¹Ï; vgl. got. gataurþs f. ai. drÌ¥Ìtiâ f.), ÏÏαÏÏá½¹Ï, á¼ÏθαÏκα, á¼Î³Î±ÏÏÎ¹Ï aus âÏÏâ ist die Stellung des Ï in κá¿Ï θέÏÏÎ¿Ï Î´á½³ÏÏ ÏÏείÏÏ ÏθείÏÏ á¼ÏθοÏα á¼Î³Îµá½·ÏÏ verantwortlich zu machen. Doch erklären sich andere unstimmige Fälle so nicht: κάÏÏα καÏÏεÏá½¹Ï neben κÏαÏá½»Ï : äol. κÏá½³ÏÎ¿Ï ÎºÏá½³ÏÏÏν κÏαÏεÏá½¹Ï, μαÏÎ½á½±Î¼ÎµÎ½Î¿Ï usw., μάÏÏÏÏ : βÏακεá¿Î½, μάÏÏÏ Ï : μέÏμεÏÎ¿Ï Î¼á½³Ïιμνα. Man darf wohl für solche Fälle mit der Möglichkeit rechnen, dass auch ein rÌ¥, das sekundär den Akzent erhielt, zu Î±Ï wurde (â¦).
It would not be inconceivable that the reflex of *rÌ¥ depended on lexical accentuation. As a parallel case one might adduce Avestan, where the reflex of *rÌ¥ is âÉrÉâ when unaccented (e.g. YAv. mÉrÉγa- âwild animalâ, cf. Ved. mrÌ¥gá- âid.â), but âÉhrâ when accented (e.g. YAv. vÉhrka- âwolfâ, cf. Ved. vrÌ¥Ìka- âid.â).105
However, whether such a scenario offers a feasible explanation in the case of Greek must ultimately depend on the data. In the above formulation by Schwyzer, it is not indicated how the difference between καÏÏεÏá½¹Ï and κÏαÏεÏá½¹Ï, both with the same accent, is to be explained. Moreover, the widely advocated analogical explanation of καÏδία, in which âαÏâ would be due to the influence of etymologically related κá¿Ï, is not straightforward either (see chapter 6). These and similar problems arouse suspicion as to whether an accent rule can solve the problem of the double reflex.
The view canonized in Schwyzerâs grammar goes back to Kretschmer (1892: 391â394). Kretschmerâs main argument for the accent rule were the Homeric particles á¼Ï (accented) beside ῥα (unaccented), which in his view retain the original distribution. For both particles, he started from a pre-form PIE *rÌ¥. Other examples adduced by Kretschmer include the gloss ÏÏá½±ÏÏοιΠαἱ Ïá½±Î¾ÎµÎ¹Ï Ïοῦ ÏÎ»á½µÎ¸Î¿Ï Ï âdivisions of the peopleâ (Hsch.), with retracted accent (beside the normal form ÏÏÏαÏá½¹Ï âarmyâ) and μάÏÏÏ Ï âwitnessâ, which he connected etymologically with βÏÎ±Î²Îµá½»Ï âarbiterâ. However, none of his examples is probative. Although there is no dialect indication for ÏÏá½±ÏÏοι in Hsch., the gloss must be from Cretan, where âαÏâ is the regular reflex also in unaccented position (see section 3.1) and where ÏÏαÏÏÎ¿Ï is indeed attested more or less with the meaning glossed by Hesychius. An etymological relation between βÏÎ±Î²Îµá½»Ï and μάÏÏÏ Ï cannot be maintained for apparent reasons. Moreover, a resolution of the problem of âαÏâ versus âÏαâ cannot be based on the particles á¼Ï and ῥα, if only because the reconstruction of their pre-form is uncertain (and they probably do not reflect *CLÌ¥T).106 Finally, as pointed out already by Grammont (1895: 26), Kretschmer did not consider the counterevidence to his rule. Of the counterexamples adduced by Grammont, καÏέδαÏθον âsleptâ and Ïá½³ÏαÏÏÎ¿Ï âfourthâ deserve to be mentioned; to these I would add καÏÏεÏá½¹Ï, ÏαÏÏá½»Ï, and καÏÏá½¹Ï.
In more recent times, Klingenschmitt (1974: 275â276) has tried to revive Kretschmerâs idea. This attempt is often cited with approval, but as I will argue in section 2.5.3, Klingenschmittâs patchy argumentation does not withstand closer scrutiny.
1.4.3 Liquid Metathesis
Since Kretschmerâs accent rule does not account for all instances of âαÏâ < *rÌ¥, some scholars have invoked liquid metathesis as an additional mechanism. Hirt (1901: 232â238) argued as follows. On Crete, âÏοâ appears to have undergone metathesis to âοÏâ in the forms ÏοÏÏι (Hom. ÏÏοÏá½·) and ÎÏοÏδιÏα (á¼ÏÏοδίÏη). Therefore, forms with âαÏâ (frequently found on Crete) need not directly continue *rÌ¥ but could also be due to metathesis from âÏαâ (assuming that this was the regular outcome). Starting from this observation, Hirt proposed that metathesized forms with âαÏâ may also occasionally appear in other dialectal areas, and were even utilized in Epic Greek because of their metrical utility. Another much-cited treatment of the problem is Lejeune (1972: 196), whose reasoning is similar to that of Hirt. Assuming that Ïα is the regular reflex and Î±Ï is analogical, Lejeune invokes the âmobilité générale des liquides dans le syllabeâ in order to account for the problematic forms that remain, such as καÏÏεÏá½¹Ï ~ κÏαÏεÏá½¹Ï.
However, forms like δÏÎ¿Î¼Î¿Ï âtrackâ and κÏÎ¿Î½Î¿Ï âtimeâ (= Att. ÏÏόνοÏ) are also attested on Crete, constituting counterexamples to Hirtâs scenario. For this reason, Hirt assumes that the supposed liquid metathesis operated on an irregular basis. As I will argue in section 3.1, a completely different scenario is more plausible: Cretan ÎÏοÏδιÏα and ÏοÏÏι may have the regular reflex of *rÌ¥ after a preceding labial consonant. Moreover, given that âÏοâ was retained in Cretan in δÏÎ¿Î¼Î¿Ï and κÏονοÏ, it is much more attractive to analyze âαÏâ as the regular Cretan reflex (at least in non-labial environments). If this is correct, it refutes the idea of a regular liquid metathesis in Cretan, and it deprives the assumption of an incidental liquid metathesis in other dialects of a solid parallel.107
Against the suggestion that liquid metathesis may operate irregularly, it must be stressed that this phonetic development in fact often operates in a completely regular and predictable way, for instance in the so-called polnoglasie forms in Slavic languages. The phonetic conditions for liquid metathesis may be highly specific: a noteworthy example is the regular metathesis of unaccented *Ér to rÉ in Le Havre French (Blevins & Garrett 1998), which seems to be conditioned by a following labial fricative or labial nasal. Thus, to assume an irregular liquid metathesis does not account for the difference âÏαâ versus âαÏâ: it merely amounts to admitting that one is unable to indicate a historical condition for the attested distributions.108 In the course of this book, we will see that the situation in Mycenaean, Homeric and Classical Greek is not so hopeless as to call for such a resignation.
1.4.4 Secondary Ablaut TeRTâ : TaRTâ
In his discussions of Indo-European ablaut, KuryÅowicz has suggested that in what he called the âSouthernâ Indo-European languages (comprising the branches of Greek, Italic and Celtic), a secondary zero grade *TaRTâ could be introduced, replacing forms of the structure *TRÌ¥Tâ before a vowel. An example from Latin is carpÅ âto pluckâ (root PIE *kerpâ), where xcorpÅ would be the expected outcome of *krÌ¥p-e/oâ. KuryÅowicz (1968: 243) proposes the following scenario. The disappearance of laryngeals in roots of the structure TeRHâ in Celtic, Latin and Greek led to the emergence of an ablaut pattern TeR-Vâ : TaR-Vâ < TRÌ¥H-Vâ in forms where the zero grade was followed by a vowel. This pattern was then analogically transferred to roots of the structure TeRTâ, yielding secondary ablaut TeRTâ : TaRTâ in cases where the suffix started with a vowel. This would explain why we find secondary zero grades like carpÅ mainly with roots of the structure TeRTâ and only rarely with roots of the structure TReTâ, where there was no corresponding model of the type TReâ : TRaâ.
KuryÅowiczâs scenario has been embraced by various scholars, including GarcÃa Ramón (1985) and Hajnal (1997: 146â150). It is problematic, however, that the evidence for secondary zero grades of the type *TRÌ¥Tâ â *TaRTâ is not at all widespread across the Indo-European realm. All nine roots adduced as evidence by KuryÅowicz (1968: 243â244) have a Latin example with a-vocalism, and in at least seven of these cases the Latin forms are the only reason to posit a secondary zero grade *TaRTâ. Indeed, the Latin a-vocalism remains problematic,109 but this is a problem belonging to the prehistory of Italic: KuryÅowiczâs idea of a âSouthernâ subgroup of Indo-European cannot be upheld.110
Further problems arise once the actual Greek evidence for secondary ablaut is considered. To be sure, there are well-known cases of secondary zero grades in Greek: for instance, the intransitive aorist á¼ÏÏάγην (á¿¥á½µÎ³Î½Ï Î¼Î¹ âto breakâ, PIE root *u̯reh1ǵâ) replaces the expected form *á¼ÏÏήγην < *e-u̯rh1ǵ-eh1â by analogy with cases like á¼Ïάγην (Ïá½µÎ³Î½Ï Î¼Î¹ âto fix; make solidâ, PIE root *peh2ǵâ).111 However, in such cases we are dealing with the extension of already-existing ablaut patterns, not with the special creation of a morphologically conditioned reflex. I do not think that the Greek forms adduced by KuryÅowicz require such a drastic explanation.
For *TaNTâ replacing *TNÌ¥Tâ, KuryÅowicz cites only two examples: κάνδαÏοÏÎ á¼Î½Î¸Ïαξ âpiece of charcoalâ (Hsch.), which is supposed to be related to Lat. candeÅ âto shineâ, Ved. cand âid.â, and Ïκάνδαλον âtrap; outrageâ (LXX+), which would be related to Lat. scandÅ âto rise, ascendâ, Ved. skand âto leapâ. It is suspicious, however, that neither κάνδαÏÎ¿Ï nor Ïκάνδαλον has an inner-Greek cognate formation with full grade root. In my view, neither etymology is compelling. Concerning κάνδαÏοÏ, the root PIE *kendâ âto shineâ qualifies bright, white light, especially that of the moon. Now, charcoal (á¼Î½Î¸Ïαξ) may glow, but it does not shine, and arguably blackness is a more specific characteristic of charcoal. As for Ïκάνδαλον, although the derivative ÏκανδάληθÏον âcurved piece of wood in a trapâ (Ar.+) assures the existence of this word for the classical period (cf. DELG s.v.), the derivation from *skendâ, accepted by both GEW and DELG, is not evident semantically.112 Given its specialized technical meaning and the a-vocalism of the root, Ïκάνδαλον may well be a loanword, for instance from Pre-Greek (cf. EDG s.v.).
For secondary *TarTâ, the only Greek form mentioned by KuryÅowicz is á¼ ÏÏη âsickleâ, related to OCS srÑpÑ, Latv. sirpis.113 However, the Greek and Balto-Slavic forms can be explained as reflexes of a root noun *srÌ¥pâ: see section 9.6. GarcÃa Ramón (1985: 217â218) has proposed to extend KuryÅowiczâs explanation of Lat. carpÅ to Greek καÏÏá½¹Ï âfruit; harvestâ. KuryÅowiczâ original idea was that Lat. sarpiÅ âto prune (the vine)â and carpÅ âto pluckâ contained analogical prevocalic zero grades *TaRT-Vâ of late-PIE date. This is, however, not the only possibility. For instance, Schrijver (1991: 493) has proposed that an a-vowel in Latin may have arisen in positions where it stood before three consonants, as in sarptus < *srÌ¥p-toâ, carptus < *krÌ¥p-toâ; subsequently the vocalism would have spread to other forms in the paradigm. Alternatively, Schrijver assumes that the âaâ may have been taken from the semantically and formally close verb sarriÅ âto hoe, weedâ.
However this may be, the most important objection to KuryÅowiczâs secondary ablaut remains that there is no obvious motivation for the assumed analogical introduction of TaRT as long as *TRÌ¥Tâ was still analyzable as a regular zero grade. His supposition that *TRÌ¥Tâ was felt to be ambiguous between *TReT-roots and *TeRT-roots does not seem a sufficient motive to me. Thus, both the lack of absolutely compelling evidence and the absence of a clear motivation for the allegedly âmorphologically conditionedâ analogy are reasons to reject the concept of secondary ablaut in Greek forms with âαÏâ.114
1.4.5 Conditioning by Neighboring Consonant Clusters
In his contribution to the second volume of Morphologische Untersuchungen, Osthoff (1879: 144â145) remarked that the outcome Î±Ï < *rÌ¥ in Greek cannot always be understood as analogical:
Es gibt fälle, in welchen man dem Î±Ï = á¹ schwerlich mit irgend welchem âsystemzwangeâ wird beikommen können. Bei καÏδία neben κÏαδίη, á¼Î´Î±Ïθον neben á¼Î´Ïαθον (â¦) und wol noch in anderen fällen fehlt uns im griechischen jegliche spur einer anderen, stärkeren ablautsstufe derselben wurzeln. (â¦) Vollends bei κάÏÏοÏ, καÏÏεÏá½¹Ï und κÏá½±ÏοÏ, κÏαÏεÏá½¹Ï, κÏαÏá½»Ï würde uns die zuhilfenahme von κÏá½³ÏÏÏν (ion.), κÏá½³ÏÎ¿Ï (lesb.), Τιμο-κÏá½³ÏÎ·Ï allenfalls nur zu dem nicht gesuchten entgegengesetzten resultat führen können, dass Î±Ï lautgesetzmässig und Ïα durch die analogie bewirkt sei. Und aus demselben grunde würden die doch nur zu ÏÏá½³ÏÏ âdick werden lassen, gerinnen machenâ unmittelbar gehörenden ÏαÏÏá½³ÎµÏ âdichtâ, Ïá½±ÏÏÎ¿Ï âdickichtâ unbegreiflich bleiben.
In order to resolve this problem, Osthoff proposed that the coda of the preceding word could influence the development of *r̥:
Hiess es ursprünglich ἡ κÏαδία mit κÏαâ im anschluss an das vocalisch auslautende, aber Ïá¿Ï καÏÎ´á½·Î±Ï mit καÏâ hinter dem consonantisch schliessenden proklitikon?
One drawback of this hypothesis is that it cannot be tested against concrete distributions in the evidence: it merely posits the earlier, prehistoric existence of contextual sandhi treatments. Furthermore, the example adduced by Osthoff has no explanatory value, as the demonstrative á½, ἡ, Ïá½¹ had not yet been grammaticalized as a definite article when the syllabic liquids were vocalized. Finally, there are counterexamples such as ÏÏÏαÏá½¹Ï and á½ÏιόÏÏÏαÏÎ¿Ï (cf. already Kretschmer 1892: 391).
In the twentieth century, a solution along the same lines was attempted by Hoenigswald. He formulated his idea as follows (Hoenigswald 1968: 22):115
the element of syllabicity which we have symbolized by [Ñ] crops up, with some phonetically recognizable effect in the daughter languages, after every two consonants not separated by a phonemic vowel ([..CCÑCCÑC..]).
In this way, two allophones of the syllabic liquids would have come into being: [LÑ] after a single consonant (or light syllable), and [ÑL] after a double consonant (or heavy syllable). Subsequently,
the post-light allophone merges with the consonant-vowel sequence Ïα (Ïο), while the post-heavy allophone merges with the vowel-consonant sequence Î±Ï (οÏ), thereby becoming prosodically long.
This formulation has some plausibility in abstract phonetic terms (note that the anaptyxis in *CRH, which took place before Proto-Greek, also depends on whether it is followed by a consonant or a vowel). However, like Osthoff before him, Hoenigswald never seriously considered the counterevidence to his claims. His scenario thus remains a paper exercise in phonetics and phonology.116
The idea of a special reflex âαÏâ after a heavy initial cluster was advocated also by Lubotsky (1994), in a discussion of the reconstruction of Ïá½±Ïξ âmeatâ. In his view, Ïá½±Ïξ regularly derives from a non-ablauting zero grade root PIE *tu̯r̥ḱâ. The shape of word-initial clusters would have automatically conditioned the vocalization: ÏÏá½±Ïξαν (3pl. aor.) âthey wrappedâ and ÏÏá½±Ïγανα âswaddling-clothesâ would display the regular reflex, while an initial cluster ÏÏÏâ (unattested in Greek) was impossible, according to Lubotsky. He also mentions the forms á¼ÏÏαλÏοÏ, ÏκαλμόÏ, and ÏθάÏμα. However, the suggestion that onsets such as /spr/, /spl/, /sphr/ were not allowed when the syllabic liquids were vocalized lacks a clear motivation: the onset cluster /str/ is not problematic at all (cf. ÏÏÏÏÏá½¹Ï âspread outâ < PIE *strh3-tóâ, which existed in the same form already in Proto-Greek when the laryngeals were vocalized); an onset /skl/ is found in ÏκληÏá½¹Ï âwitheredâ (probably reflecting *sklh1-róâ with a zero grade root); and /spl/ appears in ÏÏλάγÏνα âintestinesâ and ÏÏλήν âspleenâ.117 Thus, there appears to be no particular phonotactic reason as to why *sprÌ¥Câ had to be vocalized as ÏÏαÏCâ rather than ÏÏÏαCâ.118
In sum, there is no sufficient reason to suppose that the dialectal vocalization of *r̥ in Greek depended on the number of preceding or following consonants. Note that this may have been different for *CRHC in Proto-Greek: see section 1.2.1 and Van Beek 2021b.
1.5 Accounting for *rÌ¥ > âÏαâ
As we have seen, previous scholars have applied almost the entire linguistic toolkit to the problem of âαÏâ versus âÏαâ in Ionic-Attic, but without being able to explain all the attested forms. Within the framework of a regular change to âÏαâ, it does not appear to be possible to account for forms like καÏÏεÏá½¹Ï, κάÏÏα, καÏÏá½¹Ï and ÏαÏÏá½»Ï. I therefore hypothesize that these forms with âαÏâ are what they look like: the outcome of a regular sound change *rÌ¥ > âαÏâ in Proto-Ionic. We will encounter more evidence for this reflex along the road: see section 9.6 and, for an overview, section 12.5.
This means, however, that an account will have to be given of all forms with âÏαâ < *rÌ¥. This is the main task of chapters 4, 5, 6 and 8. In the present section, I set out the main lines of my argumentation, anticipating some of the conclusions to be reached.
1.5.1 Distributions and a New Scenario
Let us start with some remarkable distributions (discussed in more detail in section 6.1):
-
A clear majority of the forms with âÏαâ < *rÌ¥ are attested in Epic Greek, e.g. á¼Î´Ïακον, á¼ÏÏαθον, κÏαδίη, κÏαÏεÏá½¹Ï. What is more, most such forms are all but restricted to Epic Greek.
-
A number of forms with âÏαâ < *rÌ¥ have a corresponding by-form with âαÏâ that is attested in Attic and/or Ionic prose: cf. Att. καÏέδαÏθον âfell asleepâ ~ Hom. καÏέδÏαθον âid.â, Ion.-Att. καÏÏεÏá½¹Ï ~ poetic κÏαÏεÏá½¹Ï, Ion. καÏδίη ~ Hom. κÏαδίη.
-
In some cases, the âαÏâ variant is attested exclusively in prose (e.g. καÏέδαÏθον), in other cases the variants with âαÏâ and âÏαâ are both found in Homer (e.g. καÏÏεÏá½¹Ï ~ κÏαÏεÏá½¹Ï, καÏδίη ~ κÏαδίη).
In view of these distributions, it is worth investigating the possibility that the reflex âÏαâ originated in the language of epic.
Before further following this line of thought, let us pause and inquire into the reasons for previous scholars to interpret the reflex âÏαâ as a retained phonological archaism of Ionic-Attic. One motivation was, surely, that forms with âÏαâ often have earlier attestations or more archaic phonology and/or morphology. For example, the regular aorist of Ïá½³ÏÏομαι âto enjoyâ in Homer is ÏαÏÏá¿Î½Î±Î¹, but the 1pl. subj. ÏÏαÏείομεν âlet us get satisfactionâ is also attested in a formulaic verse in Homer. Whereas the root vocalism of ÏαÏÏá¿Î½Î±Î¹ may obviously have been influenced by the full grade present stem Ïá½³ÏÏομαι, the irregular form ÏÏαÏείομεν looks like a phonological archaism of Ionic that was retained because of its metrical utility.
This account of ÏÏαÏείομεν may seem plausible at first sight, but as we will see in section 6.8.5, it leads to various problems. Besides, a doublet like κÏαÏεÏá½¹Ï ~ καÏÏεÏá½¹Ï does not admit of a similar explanation because καÏÏεÏá½¹Ï cannot be analogical. In chapter 5, I will show in detail how the variation between κÏαÏâ and καÏÏâ came into being, and how it spread within the epic language by analogical mechanisms. Moreover, the common assumption that καÏδίη (~ κÏαδίη) was analogically reshaped after κá¿Ï is also highly problematic, as I will argue in section 6.1. The same problem applies to Ïá½³ÏαÏÏÎ¿Ï (~ Ïá½³ÏÏαÏοÏ), which is usually assumed to have analogically acquired the vocalism of the cardinal Ïá½³ÏÏαÏεÏ, but not its geminated consonant (see section 2.6).
One might ask whether it isnât far-fetched to posit a special epic reflex âÏαâ. In my view, it isnât. First of all, the Homeric Kunstsprache abounds in artificial formations whose creation was induced by metrical factors or the peculiarities of verse composition.119 Against this background (cf. section 1.5.2), I suggest that Homeric forms like ÏÏαÏείομεν have an artificial reflex âÏαâ < *rÌ¥. This idea gains plausibility not only from the high concentration of forms with âÏαâ < *rÌ¥ in Homer and early epic texts, and more generally from the distributions just mentioned, but also from the metrical peculiarities displayed by various such forms. Words like ÏÏá½±Ïεζα âtableâ and δÏάκÏν âsnakeâ regularly undergo muta cum liquida scansion, i.e. the sequence of plosive plus liquid does not close the preceding syllable. As we shall see in a detailed treatment of the Homeric material in chapter 6, this license is frequent in words reflecting *rÌ¥, but otherwise relatively uncommon. The same phenomenon is found in certain words with âÏοâ from *rÌ¥, such as the dat. pl. βÏοÏοá¿Ïι âmortalsâ. Wathelet (1966) therefore accounted for such scansions by assuming that they originated with the vocalization of *rÌ¥ (which first yielded âÏοâ in Mycenaean and/or Aeolic, which was later replaced with Ionic âÏαâ in certain cases). In his view, the peculiar scansion was originally preserved in formulae, but later on the license acquired a somewhat wider currency, when it was gradually extended beyond the group of words with âÏαâ or âÏοâ reflecting *rÌ¥.
These conclusions were at first fairly broadly accepted, but there has also been criticism and dissent. The problematic aspects of Watheletâs argumentation will be further examined in chapter 6. For the time being, the evidence for muta cum liquida in forms with *rÌ¥ > âÏαâ may start to appear in a different light when considered against the idea of a regular sound change *rÌ¥ > âαÏâ. Could it be that *rÌ¥ was retained âartificiallyâ for some time in the traditional language of hexameter verse after it had vocalized in the Dark Age vernaculars? If so, it is possible to view âÏαâ as a later vocalization of this retained *rÌ¥. As we will see in chapters 5 to 8, this idea is corroborated by various other metrical peculiarities, including Hoenigswaldâs discovery that the double onset consonants of κÏαδίη are never used to generate position length in the Iliad. In other words, from a prosodic viewpoint, κÏαδίη appears to behave as if the underlying phonological form was still /krÌ¥diÄ/. Another issue that is explained in the new framework is the short scansion of the first syllable of the controversial form á¼Î½Î´ÏοÏá¿Ïα (see section 1.5.3).
In view of the above arguments, I put forward the following scenario:
-
*rÌ¥ regularly developed to âαÏâ (âÏαâ by analogy) in Proto-Ionic;
-
*rÌ¥ was retained in Epic Greek at this point (it does not matter which dialect, Mycenaean, Aeolic or Ionic, was the âepic defaultâ at this stage);
-
At a much later date, this retained *rÌ¥ developed to âÏαâ, and probably to âÏοâ after a labial consonant, within the epic language;
-
Forms with âαÏâ (and with analogical âÏαâ) were introduced into Epic Greek from the Ionic vernacular.
Within this new framework, a number of pieces suddenly fall into place. First of all, *rÌ¥ had almost certainly vocalized already in Proto-Ionic and Proto-Aeolic, i.e. in the 11th century or even before that.120 Assuming that the Iliad and Odyssey were composed somewhere around 700â¯BCE,121 the retention of prosodic traces of *rÌ¥ in Epic Greek is not a trivial assumption to make. If, on the other hand, *rÌ¥ was retained in the epic Kunstsprache until not too long (around a century) before Homer, this would immediately explain why prosodic traces of *rÌ¥ are still relatively frequent in the Homeric epics.
Secondly, and perhaps even more importantly, a solution for the problem of the double reflex âαÏâ versus âÏαâ comes within reach. This requires that we can give a convincing explanation for all forms with âÏαâ < *rÌ¥ that occur outside of Epic Greek, and that we can plausibly argue that Homeric forms with âÏαâ < *rÌ¥ are traditional elements of Epic Greek. These arguments involve digressions about subtle details, such as the lexical differences between the epic Kunstsprache and the vernaculars, the prehistory of the formulaic language, metrical lengthening, or the metrical behavior of certain prosodic word-types in Epic Greek.
A third potential advantage of this new scenario is that epic forms with âÏοâ, especially those like βÏοÏοá¿Ïι displaying muta cum liquida scansion, can now be viewed as the outcome of a conditioned development, rather than as Aeolic forms. If we accept the broadly-shared assumption that forms like βÏοÏοá¿Ïι are Aeolic elements of the tradition, we are still left with the question how their pervasive muta cum liquida scansion can be explained, given that this license was highly exceptional in the Lesbian poets. Within a framework accepting the existence of an Aeolic phase, one could suppose that development 3 (the vocalization of retained *rÌ¥) took place at that Aeolic stage, or at the transition from an Aeolic to an Ionic phase. In a diffusionist framework, one could assume that development 3 took place in both parallel traditions: âÏοâ would be the reflex of retained *rÌ¥ in the Aeolic tradition, âÏαâ the reflex in the Ionic tradition. In this work, I will not make a choice between the two main competing scenarios concerning the genesis of the dialectal components of the epic tradition.122 Instead, I advocate the possibility that most epic forms with âÏοâ < *rÌ¥ are merely Aeolic in appearance: in chapter 7 I argue that they arose by a conditioned development, reflecting retained *rÌ¥ after labial consonants.
1.5.2 Epic Greek versus Vernacular Dialects
Let us now briefly recapitulate the aspects in which Epic Greek was different from the vernacular dialects, including varieties of Ionic and Aeolic.123
Epic Greek is the language of various sorts of poetry that were composed in hexameters. The prehistory of this language is the topic of fierce debates, but the following points are broadly shared among scholars:124
-
Hexameter verse was used by oral poets to compose texts in various different genres and subject-types. These traditional genres include at least heroic poetry (remembering the deeds of men past) and catalogues (genealogy).125 This manner of composition was used in extempore performance and facilitated the memorization and transmission of traditional knowledge.126
-
The dominant dialectal element of Homeric language, as of most subsequent hexameter texts, is Ionic. However, this predominance may be of relatively recent date, as indicated by the presence of forms and morphemes that can never have existed in any pre-stage of Ionic.
-
The non-Ionic elements were preserved (or adopted) because they proved useful in verse composition. A large number of these elements are archaisms that cannot be ascribed to any particular Greek dialect (for instance μέÏÏÎ¿Ï âmiddleâ). Some elements have phonological or morphological innovations that single them out as Aeolic (e.g. á¼ÏÎ³ÎµÎ½Î½á½¹Ï âwhiteâ, with a geminate reflex of *âhnâ), others can be assigned with some degree of probability to Mycenaean / âAchaeanâ (e.g. Î»Î±á½¹Ï âarmy, peopleâ),127 though ascertained instances of âAchaeanâ forms are more difficult to find in view of the higher time depth and the deficient orthography of Linear B.
-
The language has been adapted to verse composition in hexameters also by the creation of artificial forms, which arose by analogy (e.g. non-Ionic and non-Aeolic á¼Î½ νήεÏÏι âat the shipsâ), by artificially stretching up the use of an existing form (e.g. using a metrically convenient middle form instead of its active counterpart), by changing the declension class of a form (e.g. forms of ἡνιοÏÎµá½»Ï instead of ἡνίοÏÎ¿Ï âcharioteerâ), etc.128
As is well-known, there are numerous lexical differences between the language of epic and that of the classical prose authors, or even post-Homeric poetry. There is a body of words, epithets and phrases that are used exclusively by Homer. In some such cases, one may suspect that the element lost currency in the vernaculars during the two or three centuries that separate Homer from the classical period (e.g. lexical replacements, semantic developments), but in other cases the difference in register must go back to prehistoric times.
To give an example, the normal word for âman, human beingâ is á¼Î½Î¸ÏÏÏÎ¿Ï in Classical Greek. This lexeme is used frequently already in Homer, and it is also found in Linear B as a-to-ro-qo. In addition, however, Epic Greek uses another form βÏοÏá½¹Ï, etymologically meaning âmortalâ but often used as a synonym of âman, human beingâ. The form is never used by later prose authors, except when they imitate Homer or attempt to write in an elevated style. We can be confident that this word was not used in everyday Ionic or Attic speech, and that it is a traditional element of poetic diction. In this particular case, we are helped by historical phonology: the form βÏοÏá½¹Ï must have developed from *mrÌ¥tós, but âÏοâ cannot be the regular Ionic-Attic reflex of the syllabic liquid. Similar arguments can be adduced for various other words or word-forms that are used predominantly in Epic Greek, especially in cases where we have reason to assume that we are dealing with an artificial formation. I will therefore regularly make use of a distinction between the Ionic-Attic vernacular and Epic Greek in what follows, even if this distinction necessarily becomes more fluid as we move back in time from the classical period towards Homer and further back.
A distinction between vernacular dialects and epic register is regularly made by scholars dealing with the artificial nature of Homeric language. Following Milman Parry, it is normally assumed that Epic Greek underwent the linguistic changes of the poetsâ vernacular, except in the case of forms that were formulaic or metrically protected in some other way.129 The above scenario, however, assumes a prolonged retention of the sound *rÌ¥ in Epic Greek. This can only be imagined if Epic Greek was a separate register, with not only its own morphology, syntax and lexicon (as is generally admitted), but also with a proper phonology and phonetics. Thus far, however, no instances of artificial phonology have been identified. This is surely due in large part to the fact that we only have a written text of the Homeric epics (which makes it difficult to say anything about phonetic realizations), and that the orthography of this text partly reflects spelling practices of the 4th century BCE. However, it is almost inevitable that epic poets would have avoided an all-too-local pronunciation in their performances, and it is plausible that certain phonetic or phonological features of the traditional poetic language were supra-regional.130 It is even conceivable that a sound like *rÌ¥, when it was progressively eliminated from vernacular dialects, came to be perceived as a marker of traditional, elevated epic style. Of course, this is mere speculation, but the point is that the scenario proposed here is not excluded by what we know about the language of early Greek epic.
1.5.3 Metrical Irregularities and the Prehistory of the Hexameter
A final issue that must be briefly addressed is the antiquity of the hexameter. Since Nagy (1974) and especially Berg (1978), various prominent Indo-Europeanists have subscribed to the idea that the hexameter arose from metrical cola inherited from Proto-Indo-European. Before that, since Meister (1921: 58) and Meillet (1923: 60â¯ff.) there was some sort of consensus that the hexameter was borrowed as a whole from the Minoans. The origin of the hexameter is an extremely difficult issue to resolve because there is little concrete evidence, nor a clear framework in which to interpret this evidence. In my view, the hexameter may ultimately derive from inherited meters, but it may also be that we lack the means to prove this, due to the antiquity of the tradition.131
At least since the early nineteenth century, scholars have used systematically occurring metrical anomalies for reconstructing earlier linguistic forms of the epic language. A clear example is the loss of word-initial *u̯â in pre-Homeric Ionic, which explains the fact that words with etymological *u̯â are more often involved in hiatus and irregular position length than one would expect on average. More controversial is the idea that metrical irregularities in words with Ïο and Ïα can be eliminated by tracing them back to a pre-form with *rÌ¥. Concerning the best-known instance, the verse-end á¼Î½Î´ÏοÏá¿Ïα καὶ ἥβην, Wackernagel (1916: 172) already remarked that the scansion of á¼Î½Î´ÏοÏá¿Ïα can be understood if the original form had *anrÌ¥â. In this he has been followed by Mühlestein (1958: 224 n. 20), Ruijgh, Wathelet and many later scholars.
However, this explanation was called into question by Tichy (1981), who argued that the scansion of á¼Î½Î´ÏοÏá¿Ïα is to be explained as a metrical archaism: the form would be a relic from Bergâs proto-hexameter, at a stage which allowed for a trochaic fourth foot. Various Indo-Europeanists have since expressed their support for Tichyâs scenario.132 At the same time they criticize the alternative viewpoint (which views á¼Î½Î´ÏοÏá¿Ïα as a phonological archaism): it would dogmatically take for granted the antiquity of the hexameter, without adducing independent proof for this claim.
Against these strong assertions of belief in the Berg-Tichy scenario, it must be stated very clearly that none of its advocates has been able to identify a means of testing it against competing scenarios. One clear criterion would be the scenarioâs ability to account for metrical irregularities and unexpected morphology that cannot be accounted for in other ways. In reality, however, Bergâs proto-hexameter (as applied by Tichy to aberrant Homeric scansions) runs a heavy risk of becoming circular, as it does not explain much more than irregularities that can also be due to prehistoric sound changes, like the vocalization of *rÌ¥.133 Apart from the much-discussed verse-end á¼Î½Î´ÏοÏá¿Ïα καὶ ἥβην, there is no evidence for the assumed trochaic fourth foot that cannot be explained otherwise. In order to show how patchy the evidence is, let us discuss in more detail a few forms that have been adduced.
Berg and Lindeman (1993: 186â193) analyze dactylic forms of the stem á¼Î½á½³Ïâ (with metrically lengthened á¼â),134 which are frequent in the fourth foot, as artificial stretched forms replacing original trochaic ones with á¼Î½Î´Ïâ.135 Thus, the words á¼Î½á½³ÏÎµÏ á¼Ïθλοὶ á½ÏονÏο occurring at the end of the line (Od. 3.471; á½ÏονÏαι Od. 14.104) would be a transformation of *á¼Î½Î´ÏÎµÏ á¼Ïθλοὶ á½ÏονÏο, which they view as an old âpherecratean formulaâ (1993: 193). They reject the scenario already proposed by Schulze (1891), who argued that á¼Î½á½³Ïα, á¼Î½á½³ÏεÏ, á¼Î½á½³ÏÎ±Ï and á¼Î½á½³Ïε (the original forms, later to be replaced by á¼Î½Î´Ïα, á¼Î½Î´ÏεÏ, etc.) underwent metrical lengthening. In that scenario the gen. and dat. sg. forms á¼Î½á½³ÏοÏ, á¼Î½á½³Ïι are artificial analogical creations beside the inherited forms á¼Î½Î´Ïá½¹Ï, á¼Î½Î´Ïá½·.
The objections formulated by Berg and Lindeman against Schulzeâs metrical lengthening scenario are:
-
Forms like á¼Î½á½³Ïα (ending in a vowel) could have been used in the hexameter without metrical lengthening, but they are never so used;
-
Forms like á¼Î½á½³ÏÎµÏ (ending in a consonant) could not have undergone metrical lengthening, as they could be placed before words starting in a consonant; á¼Î½á½³ÏεÏ, however, is never used in this way;
-
As early as the Mycenaean period, Greek dialects could not have preserved ablauting paradigms of the type ÏαÏá½µÏ / ÏαÏá½³Ïâ / ÏαÏÏâ, except in kinship terms.
The two objections concerning metrical lengthening are easily dismissed. First of all, Berg and Lindeman seem to have missed that all instances of original anapestic scansion (in forms of the shape á¼Î½á½³Ïα, á¼Î½á½³ÏεÏ, á¼Î½á½³ÏÎ±Ï with short á¼â) may have been replaced completely by the innovative forms á¼Î½Î´Ïα, á¼Î½Î´ÏεÏ, á¼Î½Î´ÏÎ±Ï once these had ousted the older forms from the spoken language. The stem á¼Î½Î´Ïâ is placed in the thesis 13â¯Ã on 72 instances of á¼Î½Î´ÏεÏ, and 10â¯Ã on 44 instances of á¼Î½Î´ÏαÏ, and this includes some archaic-looking phrases.136 As for (ii), it is true that metrical lengthening of tribrachic forms ending in a consonant (e.g. á¼Î½á½³ÏεÏ) is less frequent than with forms ending in a vowel (e.g. á¼Î½á½³Ïα), but the phenomenon does occur, especially with forms that are part of a larger phrase. A well-known example is the lengthened initial vowel of the gen. á½Î´Î±ÏοÏ, which is explained by the occurrence of this form in the phrase ΣÏÏ Î³á½¸Ï á½Î´Î±ÏοÏ. Indeed, the á¼Î½á½³Ïâ forms also occur more than once in larger phrases: cf. verse-final á¼Î½á½³ÏÎµÏ á¼±ÏÏοκοÏÏ ÏÏαί, á¼Î½á½³ÏÎµÏ á¼Î³ÏοιῶÏαι, and the frequent verse-initial á¼Î½á½³ÏÎµÏ á¼ÏÏε, Ïίλοι âbe men, friends!â. Objection (iii) is not cogent, as á¼Î½á½µÏ is (just like ÏαÏá½µÏ) a high-frequency item in which an archaic inflection may well have been preserved longer.
The speculation that verse-final á¼Î½á½³ÏÎµÏ á¼Ïθλοὶ á½ÏονÏâ arose by a transformation of *á¼Î½Î´ÏÎµÏ á¼Ïθλοὶ á½ÏονÏâ is quite bizarre. In reality, in both attestations the entire formulaic phrase stretches from |P until the end of the line: á¼Ïὶ δâ á¼Î½á½³ÏÎµÏ á¼Ïθλοὶ á½ÏονÏο âand noble men watched over itâ (Od. 3.471), á¼Ïὶ δâ á¼Î½á½³ÏÎµÏ á¼Ïθλοὶ á½ÏονÏαι (Od. 14.104), with the preverb in tmesis. In other words, there never was a âpherecratean formulaâ *á¼Î½Î´ÏÎµÏ á¼Ïθλοὶ á½ÏονÏο. In reality, á¼Ïá½· ⦠á½ÏονÏο is a clear reminiscence of Mycenaean o-pi , qe-to-ro-po-pi , o-ro-me-no (PY Ae 134), and the specific combination with the preverb may well be a phraseological relic from the Mycenaean period (as Hajnal 1998: 48 rightly notes). It is therefore highly plausible that the phrase á¼Ïὶ δâ á¼Î½á½³ÏÎµÏ á¼Ïθλοὶ á½ÏονÏο contains a metrically lengthened relic form á¼Î½á½³ÏεÏ.
From an Indo-Europeanist perspective, the forms á¼Î½á½³Ïα, á¼Î½á½³ÏεÏ, á¼Î½á½³ÏÎ±Ï and á¼Î½á½³Ïε clearly reflect the expected full-grade stem *h2nérâ of the strong case forms.137 The form á¼Î½á½³ÏÎµÏ is also the most frequent of all the á¼Î½á½³Ïâ forms in Homer and it occurs, as we saw, in formulaic phrases. Thus, dactylic forms like á¼Î½á½³ÏÎµÏ may have arisen by metrical lengthening of the first syllable of a tribrachic form /aneres/, as long as that form was available. The gen. and dat. sg. forms á¼Î½á½³ÏοÏ, á¼Î½á½³Ïι arose by an artificial extension of the pattern seen in á¼Î½á½³ÏεÏ, á¼Î½á½³ÏÎ±Ï beside á¼Î½Î´ÏεÏ, á¼Î½Î´ÏÎ±Ï after forms of the latter type had come into being. When unlengthened anapestic forms like /aneres/ were no longer current in the spoken language, they were replaced by á¼Î½Î´ÏεÏ.
In another contribution, Berg and Haug (2000) assume that the case forms νέεÏ, Î½á½³Î±Ï and Î½Îµá½¹Ï of the word for âshipâ in the fourth foot were substituted for trochaic *νá¿ÎµÏ *νá¿Î±Ï *Î½Î·á½¹Ï when these were followed by the formulaic epithet á¼Î¼ÏιέλιÏÏα. This is an unnecessary assumption: the low numbers of attestations of Î½á½³ÎµÏ and Î½á½³Î±Ï (in comparison with the numbers for νá¿ÎµÏ and νá¿Î±Ï) are compatible with the relatively recent spread of a linguistic innovation (cf. Hoekstra 1965: 124â130).138 I would like to add that the original form of this formula may have been the acc. pl. in the form *Î½Î±á¿¦Ï á¼Î¼ÏιελίÏÏÎ±Ï (with Î½Î±á¿¦Ï as attested in Attic, which may be an archaism).139
As a third example, Hajnal (2003: 76 n. 124) refers to the odd 3pl. form μιάνθην in the verse-end μιάνθην αἵμαÏι μηÏοί (Il. 4.146) for expected *μίανθεν, suggesting that the latter form (which would be expected in Bergâs proto-hexameter) was actually sung by the original Iliad poet. However, although μιάνθην is indeed a strange form, it cannot be excluded that it was a one-off analogical creation (cf. Meister 1921: 25).
Scholars have also adduced artificially lengthened forms such as ÏÏολιÏá½¹ÏθιοÏ, á¼á½³Î¸Î»Î¹Î± (for expected ÏÏολίÏοÏθοÏ, á¼ÎµÎ¸Î»Î±) as evidence for Bergâs proto-hexameter, but these forms could equally well be accounted for in another framework, such as that of Witte (see below); they do not necessarily imply the existence of an earlier verse-form with a trochaic-ending fourth foot.
In sum, the purely hypothetical character of Bergâs scenario appears, first of all, from the fact that no less than four âtransformationsâ (cf. the clear summary in Hajnal 2003: 74â75) are needed to reach the attested hexameter from the putative starting point. Secondly, it explains only the genesis of the hephthemimeral caesura, not that of the more important bucolic dieresis and the third foot caesuras. Thirdly, assuming trochaic-ending pre-forms does not offer a convincing explanation for the peculiar linguistic forms occurring in the fourth foot. And finally, even if a scenario like that of Berg were correct, we have no idea at all when the hexameter would have come into being.140
In my view, Bergâs scenario (and similar ones) are clearly inferior to an almost forgotten proposal by Witte (1913), who argued extensively for deriving the hexameter from a combination of a dactylic tetrameter plus an adoneus.141 It has the advantage of accounting for the high general frequency of the bucolic dieresis, and also for the fact that the bucolic dieresis is the place where clause boundaries are most frequent. Furthermore, as Witte shows, the combination of a tetrameter plus an adoneus is actually attested in Greek poetry. Also, the two metrical laws that occur in the thesis of the fourth foot, Hermannâs Bridge and Wernickeâs Law, follow more or less automatically from the scenario. On the other hand, two caveats that were mentioned above with respect to other proto-hexameter theories apply to Witteâs scenario: it is not easy to test it against the evidence, and we have no idea when the hexameter would have acquired its Homeric form.
One gets the impression that the assertions of belief in Bergâs scenario were guided by, among other things, a desire to get rid of the pre-Mycenaean origin of the tradition argued for by classicists like Ruijgh and West.142 Indeed, as we will see the linguistic arguments for such an early origin are weak. However, given that the epic tradition is highly conservative in its formulaic language, and that the preservation of archaisms in this language must be understood as a function of the systemâs thrift, it is difficult to see why the meter itself (which, after all, caused this system to develop) would not be equally conservative. Of course, the formulaic language was subject to continuous updating and reworking, as scholars like Hainsworth (1968) and Hoekstra (1965) have shown in detail. However, these modifications can be understood as the poetsâ response to linguistic changes (they tried to remain comprehensible), combined with an attempt to maintain or even expand the economy and thrift of their system of verse composition. Even if we take into account these modifications, it can hardly be denied (as Hoekstra 1981 has argued in detail) that the formulaic system has a traditional core that goes back generations, and which presupposes the existence of something very much like the Homeric hexameter at an early date. In the course of this book, we will encounter various indications that corroborate this conclusion.
1.6 Outlook
Leaving aside the environments (discussed in section 1.2) in which an anaptyctic vowel emerged beside *r̥ and *l̥ already in Proto-Greek, my aim is to answer the following three questions:
-
What was the regular development of Proto-Greek *r̥ and *l̥ in the major Greek dialect groups?
-
Which mechanisms affected the development of forms with etymological *r̥ in Epic Greek?
-
What can be inferred, from the vocalization of *r̥ as an isogloss, about the genesis and prehistory of the four main dialect groups, and of Epic Greek?
In view of the possibility that *r̥ and *l̥ vocalized in different ways and at different times, the evidence for *l̥ will be treated separately in chapter 10. I will start, in chapters 2 and 3, with the regular development of *r̥ in all dialects apart from Ionic-Attic and Epic Greek. This requires that all available etymological evidence is evaluated and sifted. Special emphasis will be placed in these chapters on the question concerning the regular slot in which the anaptyctic vowel developed: before or after the liquid.
The treatment of the dual Ionic-Attic reflex (âαÏâ beside âÏαâ) starts in chapter 4 with one specific morphological category, the so-called âCaland formationsâ. This portion of evidence is important in that it illustrates that the majority of forms with âαÏâ and âÏαâ can be due to analogy, and hence are not probative as regards the regular reflex of Proto-Greek *rÌ¥.
Chapter 5 is devoted to one specific set of âCalandâ formations, the etymological family of κÏαÏεÏá½¹Ï ~ καÏÏεÏá½¹Ï, in which a number of different words occur in doublets, both in the classical language and in Epic Greek. Our goal in this chapter will be to determine the linguistic processes by which the doublets originated.
The conclusions reached in chapters 4 and 5 confirm the point that both âαÏâ (ÏαÏÏá½»Ï, καÏÏεÏá½¹Ï, κάÏÏα) and âÏαâ (κÏαÏαιόÏ, θÏαÏá½»Ï) are regular reflexes of *rÌ¥. At first sight, this seems to confirm the impasse, but the analysis of θÏαÏá½»Ï versus θάÏÏÎ¿Ï in chapter 4 and that of καÏÏεÏá½¹Ï versus κÏá½±ÏÎ¿Ï in chapter 5 have another consequence: they show that the analogical developments leading to doublet forms in Epic Greek were quite different from the developments taking place in the vernaculars.
In chapter 6, a distribution is established for doublets with âÏαâ and âαÏâ: the occurrence of âÏαâ is shown to be limited to Epic Greek, while âαÏâ may occur both in Epic Greek and in classical prose. The chapter then considers in full detail all forms with âÏαâ occurring in Early Greek Epic. Many Homeric forms with âÏαâ are characterized by metrical peculiarities, notably muta cum liquida scansion. An fresh analysis of all instances of muta cum liquida in Homer confirms that the phenomenon correlates strongly with the original presence of *rÌ¥. Thus, combining the distribution of forms with âÏαâ with their metrical behavior, I hypothesize that *rÌ¥ was retained longer in the epic tradition in the way sketched in section 1.5.
This new framework is also applied to epic forms with âÏοâ: in chapter 7, I investigate the hypothesis that âÏοâ is the regular reflex of Epic *rÌ¥ after a labial consonant. This chapter also includes a discussion of the phrases á¼Î½Î´ÏοÏá¿Ïα καὶ ἥβην and á¼Î½Ï αλίῳ á¼Î½Î´ÏεÏÏόνÏá¿ that have played such an important role in previous discussions. The evidence for âαÏâ and âÏαâ in thematic aorist forms is discussed separately in chapter 8, because the metrical behavior of these formations is different from that of other forms with Epic *rÌ¥.
Chapter 9 treats several remaining loose ends. I will revisit three specific phonological environments in full detail: *ârÌ¥sâ, word-final *ârÌ¥, and *ârÌ¥nâ. Also, I discuss the more marginal evidence (uncertain and implausible etymologies) and present an overview of further Ionic-Attic evidence for a regular reflex âαÏâ < PGr. *rÌ¥.
After treating the evidence for *l̥ in chapter 10, including the dialectal reflexes and the possibility of discerning conditioned developments, chapter 11 uses the new insights concerning *r̥ and *l̥ to obtain a relative chronology. This will allow us to draw definite conclusions about the vocalization of *r̥ as an isogloss in the prehistoric development of the Greek dialect groups. In chapter 12, finally, I will take stock on the basis of a summary of my main conclusions, and ask whether the benefits of the new framework outweigh its potential drawbacks.
The PIE phonological system with both syllabic liquids and syllabic nasals is placed in a typological perspective by Cooper (2013).
Throughout this book, accent marks and breathing signs will not be added to alphabetic Greek forms as attested in inscriptions.
See, for instance, Ruijgh (1961; 1967; 1985; 1995; 1997).
âAchaeanâ is the conventional name for the hypothetical dialect group comprising Mycenaean, Arcadian and Cypriot.
See e.g. Buck (1955: 20); Lejeune (1972: 197).
North Greek comprises the later West Greek and Aeolic groups, and Proto-South Greek is the ancestor of âAchaeanâ and Proto-Ionic. The idea was already proposed before the decipherment of Linear B: see Risch (1949) and Porzig (1954). For a history of early research on the possible relations between Mycenaean and the precursors of the alphabetic dialect groups, see Cowgill (1966).
Cf. Myc. di-do-si /didonsi/ âthey giveâ (Ion.-Att. 3sg. δίδÏÏι), and Myc. to-so /to(s)son/ âso muchâ (Ion.-Att. Ïá½¹ÏοÏ), Myc. me-sa-to /me(s)sato-/ (Ion.-Att. Arc. μέÏÎ¿Ï âmiddleâ). The ambiguous spelling of Linear B does not allow us to determine whether Mycenaean had already undergone the development *âssâ > âsâ. The Boeotian and Cretan reflexes presuppose that the affricate reflecting intervocalic PGr. *âti̯â was preserved in Proto-Aeolic and Proto-West Greek, respectively.
I write /ddj/ for the outcome of the palatalized stops *di̯ and *g(w)i̯ in Mycenaean, following the arguments advanced by e.g. Risch (1979b) and Crespo (1985). Of course, my argument concerning the syllabic liquids does not depend on this; the reader who wishes to read e.g. a geminated affricate /ʣ:/ instead may feel free to do so.
Apart from Ruijgh (in most of the publications cited in n. 2), cf. e.g. Hackstein (2002: 6); Barnes (2011: 2).
In a-di-ri-ja-te /andriantÄ/ (ins. sg.) âwith a manâs figureâ, the PN a-re-ka-sa-da-ra /AleksandrÄ/, and perhaps in the PN a-da-ra-ko /Andrarkhos/. However, these forms do not contain the reflex of *rÌ¥ but of its prevocalic consonantal allophone, i.e. *anrVâ. See section 7.3.1.
Apart from Ruijgh, see also Cowgill 1966: 92â93. However, this view is certainly not shared universally: among the scholars defending a pre-Mycenaean origin of Epic Greek, West (1988: 156â157: âin the Mycenaean tablets that stage is already past; that dialect at least [emphasis LvB] has moved irrevocably towards or or roâ) and Wathelet (1970: 172: âun fait relativement récent en mycénien et, sans doute, aussi dans lââ¯ensemble du grecâ) are much more cautious.
This is specifically Ruijghâs view (e.g. 1985: 162â163, 1992: 84â87, 1996: 117). Similarly, Cowgill (1966: 94â95).
For a summary overview of morphological criteria (the athematic inf. act. in âναι, âμεν, or âμεναι, or adverbs of the type á½ Ïε, á½ Ïα, ὠκα âwhenâ) and lexical criteria (e.g. βούλομαι, βόλομαι vs. δήλομαι, βέλλομαι), see the classic study by Risch (1955), especially the table on p. 75, and also Cowgill (1966). The most important question is always whether a feature shared by two dialect groups can be proven to be a common innovation, or whether we may be dealing with shared archaisms or independent innovations.
This view is maintained also in Risch 1979a. However, Rischâs views on the syllabic sonorants are idiosyncratic in at least two respects. First, he views the vocalization of syllabic liquids and nasals as part of the same development (which is implausible: see section 1.3.3). Secondly, he seems to have held that the o-vocalic outcome in Aeolic and Arcado-Cyprian is an archaism, i.e. that it can be viewed as a pre-stage of the a-colored reflex in Ionic-Attic, which is untenable (cf. the criticism in Cowgill 1966: 80 and 82).
See chapter 6.
Mühlestein (1958: 224): âDemnach muss schon vor der Mitte des zweiten Jahrtausends in griechischen Hexametern von Mannheit gesungen worden seinâ. See also Ruijgh (as above), Wathelet (1966: 171â172), West (1988: 156â157). However, Mühlestein (1958: 226, Nachtrag) also argued that âder Weg zur homerischen Sprache (â¦) nicht durchs Mykenische hindurch, sondern am Mykenischen vorbei [geht]â in view of the abstract a-no-qa-si-ja which he interpreted as /anorkwhasiÄ-/, excluding âroâ as a regular reflex. For further evaluation of these arguments, see section 7.3.3.
Hackstein (2002: 6) speaks of âein beständiger Zankapfel zwischen der Philologie und der Sprachwissenschaftâ, which is not quite accurate because the âphilologistâ view has also been championed by linguists.
E.g. Haug (2002), Hackstein (2002; 2010), Hajnal (2003). However, note that Tichyâs monograph on the subject (2010) has been severely criticized by West (2011) in his review of it.
Cf. Haug (2002: 63â64), whose arguments concerning á¼Î½Ï αλίῳ á¼Î½Î´ÏεÏÏόνÏá¿ will be further discussed in section 7.3.2.
Cf. e.g. Hackstein (2002: 6). The absence of καί in Mycenaean caused Ruijgh to modify his views on á¼Î½Î´ÏοÏá¿Ïα καὶ ἥβην in his later publications (e.g. Ruijgh 1997: 42â44); as a consequence, he then placed more emphasis on á¼Î½Ï αλίῳ á¼Î½Î´ÏεÏÏόνÏá¿ as the main piece of evidence for the assumed pre-Mycenaean origins of epic verse.
Cf. Risch (1955: 72 and 1979a: 109) and Heubeck (1972). According to Wathelet (1970: 172â173) the vocalization âconstitue un fait relativement récent en mycénien et, sans doute aussi, dans lââ¯ensemble du grec.â
See section 3.4.
For instance, Cowgill (1966) is cited with approval by Parker (2008), and Heubeck (1972) has been accepted by GarcÃa Ramón (explicitly in 1975, implicit in many later works). Rix (1992: 65) is more reserved about the o-colored outcome in Mycenaean and Arcado-Cyprian. Lejeune (1972: 197â198) assumes a stronger âpreferenceâ for the o-colored outcome in Mycenaean and the Aeolic dialects.
Apart from Wathelet (1970) and West (1988), this thesis is accepted by scholars like Sihler (1995: 92), Haug (2002: 59), Hackstein (2002: 5â7), Hajnal (2003). Meier-Brügger (1992b: 288) and Barnes (2011: 2 with n. 6) use the form á½ÏόδÏα as an argument for an early vocalization, but as I will show in chapter 9, this is unfounded.
Cf. Haug (2002); Hackstein (2002); Hajnal (2003).
See e.g. Lejeune 1972: 197.
Since the evidence for *l̥ is too limited, I will focus on the vocalization of *r̥ as far as chronological issues are concerned. It cannot be excluded on forehand that *l̥ vocalized earlier than *r̥.
GarcÃa Ramón (1975) assumes a post-Mycenaean vocalization to âÏοâ, âοÏâ in Proto-Aeolic, basing himself on Heubeckâs idea of retained *rÌ¥ in Mycenaean. However, there is no principled reason to conclude, from the non-occurrence of a change in one (South Greek) dialect, that the change did not occur in a different (North Greek) dialect.
In this respect, I agree with Heubeck (1972).
Forms like OFris. ankel, OHG enchil âankleâ and Lat. umbilÄ«cus ânavelâ corroborate the antiquity of these formations.
On Sieversâ Law, see generally the monograph by Barber (2013) and my review of it, Van Beek (2016).
Note the following notations: R = any sonorant (= liquid or nasal), L = any liquid, N = any nasal.
Beekes (1969) at first defended the idea that the anaptyctic vowel before the sonorant in the sequence *CÉRHV could be colored by the laryngeal following it (âlaryngeal umlautâ). He gave up this idea soon afterwards in view of á¼Ïήν âlambâ beside Ved. úranâ, both reflecting PIE *u̯rh1-Än, and á¼Î»á¿¶Î½Î±Î¹ âto get caughtâ < *u̯lh3-eh1â. Ruijgh kept defending the idea of âlaryngeal umlautâ in various publications. For further arguments against it, see Peters (1980: 27â31 n. 19).
For this point, see e.g. GarcÃa Ramón (1985) and Rix (1992: 74).
For ÏόλαιÏι beside ÏαλάÏ, one might compare other Lesbian forms with ο for α where this interchange has nothing to do with the development of *CRHV, e.g. κόθαÏÎ¿Ï for καθαÏá½¹Ï âpure, cleanâ (see section 9.7.2).
The only apparent exception is West Greek ÏÏᾶÏÎ¿Ï âfirstâ corresponding to ÏÏῶÏÎ¿Ï in the other dialects (cf. Beekes 1969: 214â216). Peters (1988: 376) admits a special development of *CRHC in West Greek, but the difference ÏÏᾶÏÎ¿Ï vs. ÏÏῶÏÎ¿Ï can be explained in other ways. It has been argued that ÏÏᾶÏÎ¿Ï has the regular reflex of *prÌ¥h2-toâ, and that the vowel color of ÏÏῶÏÎ¿Ï was influenced by that of ÏÏá½¹ÏεÏÎ¿Ï âearlier, beforeâ (Rix 1992: 73, Cowgill 1966: 149); for another explanation, see Waanders 1992: 378.
Concerning this issue, cf. the informative recent discussion by Höfler 2016/17.
For two other possible instances of a disyllabic reflex /aRe/, see Höfler 2016/17.
Cf. e.g. Haug (2002: 53) following GarcÃa Ramón (1985: 206â208).
Peters 1980: 223â228 argues that ἰοÏέαιÏα contains the word for âhandâ, but doubts whether the form can reflect PGr. *isu̯o-khesrÌ¥-i̯a; for this reason, he envisages a reconstruction *âǵheserih2 (analogous to the motional feminine ÏίειÏα, Ved. pÄ«varÄ«Ì) or *âǵhesrih2, and argues that an outcome *á¼°á½¹ÏειÏα would be expected in both cases. In the end, he asks whether verse-final ἰοÏέαιÏα may be a Homeric Streckform which secondarily obtained its suffix âαιÏα from elsewhere. I find it much more likely that the form directly reflects *isu̯o-khesrÌ¥-i̯a, but the issue will not be further pursued here.
See Peters 1980: 198 and section 9.5. Aeol. á½Î½Î¿Î¹ÏÎ¿Ï has also been compared to Arm. anurÇ° âdreamâ < *onÅri̯oâ, a pre-form which would yield á½Î½Î¿Î¹ÏÎ¿Ï directly by Osthoffâs Law, but again, positing two pre-forms *onÅri̯oâ beside *oneri̯oâ would be unparalleled from a morphological (or derivational) perspective.
Moreover, ἠλλομαι may have been influenced by the Hom. root aorist á¼Î»Ïο (*ἠλÏο). Examples like ÏÎºá½±Î»Î»Ï âto hewâ and ÏÏάλλομαι âto stumbleâ can be derived from older nasal presents, and in any case their root does not undergo ablaut; they are therefore irrelevant for the present discussion.
Cf. GarcÃa Ramón (1985: 207). The reflex in Ved. háryati differs from that in mriyáte âto dieâ, which must contain the regular Indo-Aryan reflex of PIE *Cr-ie/oâ (cf. Lat. morior < *mrÌ¥-i̯e/oâ). According to de Vaan (EDL s.v. horior), this fact and the full grade root herâ attested in Sabellic point to an ablauting PIE i-present. It seems more likely to me that these full grades were introduced from elsewhere in the verbal paradigm, and that the PIE present stem already had thematic *âi̯e/oâ.
For a recent discussion of these questions of syllabification in an optimality theory framework, see Cooper 2014, chapters 6 to 10.
Cf. GarcÃa Ramón (1985: 207).
Cf. Lamberterie (1990: 187â194).
Lamberterie (1990: 192â193) proposes that *mnH-uâ âsparse, rare, thinâ contains the verbal root of Lith. mìnti âto tread, break flaxâ; if so, we might assume that an older meaning of the adjective was âbroken into piecesâ.
In section 10.5.3, I argue that ÎºÎ±Î»á½¹Ï reflects a thematicization of PGr. *kalúâ < PIE *klH-uâ, containing the root of Lith. kìlti âto rise, emergeâ, Lat. âcellÅ âto stand out, excelâ.
GarcÃa Ramón (1985: 210) also remarks that there is no good evidence for the reflex of *CrÌ¥u̯Vâ, but makes a possible reservation concerning Hom. ÏᾶÏÎ¿Ï and Myc. pa-we-a2. Since he also accepts that ÏᾶÏÎ¿Ï could be a loanword, I do not understand on which basis he concludes that âthe Common Greek form must be reconstructed as *phËrwos > *pharwosâ.
Thus already Schwyzer (1939: 342): âÎ±Ï erscheint im Anlaut und Auslaut, vor einstigem Halbvokal und vor Vokalâ. Cf. also Morpurgo Davies (1968) and GarcÃa Ramón (1985).
Cf. Ruijgh (1992: 86 n. 31).
Cf. Haug (2002: 50).
This phenomenon is generally known as âLex Rixâ (Rix 1970) and mostly thought to apply to both liquids and nasals. On *HLCâ, see also Vine (2005).
The comparison of the particle á¼Ïα âthen, thereforeâ, Hom. á¼Ï with the Baltic conjunction Lith. irÌ, Latv. ìr âand, also; evenâ and/or with the question particle Lith. arÌ, Latv. ar, has been taken to point to a reconstruction *rÌ¥. This is based, essentially, on the identification of á¼Ïα with Homeric á¼Ï, ῥα (cf. Hoenigswald 1953: 289â290, with a review of older literature). Upon this view, á¼Ïα arose as a conflation of á¼Ï and ῥα, which are both supposed to be outcomes of a pre-form *rÌ¥. Haug (2002: 52) accepts the reconstruction *rÌ¥ for the Greek forms, but admits that all kinds of special accidents may have taken place in a monosyllabic clitic, and therefore does not use á¼Ïα, á¼Ï, ῥα as evidence. I agree with this conclusion, but since the origin of the variation á¼Ï ~ ῥα remains obscure, I am not so sure about the reconstruction *rÌ¥. It is possible to connect á¼Ï with the Baltic forms if we reconstruct a particle *h2r, *h2ér (EDG s.v. á¼Ïα, cf. also DELG), but this would leave the origin of ῥα unaccounted for. Another option is to reconstruct á¼Ï, ῥα as *srÌ¥; this would explain the existence of a Mycenaean particle âa2 (in o-da-a2, o-de-qa-a2) but it does not yield a meaningful etymology.
On á¼ÏÏÎ±Î»á½³Î¿Ï < *u̯alpaléoâ (with analogical âalâ < *lÌ¥), see section 10.2.1.
See e.g. Schwyzer (1939: 342), Lejeune (1972: 196), GarcÃa Ramón (1985), and Sihler (1995: 92).
See e.g. Ruijgh (1961), Peters (1980).
âPeut-être le développement de RÌ¥ syllabique en αR ou Rα est-il grec commun non seulement devant voyelle, y et w, mais encore devant toute sonante. (â¦) on lit, à Mytilène et à Larisa, ÏÏάλλα (= att. ÏÏήλη) qui provient de *stlÌ¥-nÄâ (â¦). Il semble bien quââ¯il y ait eu développement dââ¯une voyelle de timbre a devant sonante dans ces dialectes qui attestent normalement, en position interconsonantique, un o.â
Cf. Rix 1992: 67, DELG s.v.
For the suffix *âsleh2â, one may compare e.g. Lat. scÄlae âstairsâ < pre-Latin *skand-slÄâ.
As Hirt already remarked, âdie nÄâ und neu-Verben haben αÏâ (1897: 157), mentioning as examples, among others, ÏÏá½±ÏÎ½Ï Î¼Î±Î¹ âto sneezeâ and the Hesychius gloss θάÏÎ½Ï Ïθαι (for θόÏÎ½Ï Î¼Î±Î¹ âto copulateâ).
Baderâs reference to the supposedly unconditioned double reflex of the syllabic sonorants in Balto-Slavic is erroneous, because the conditioning factor for âuRâ was a preceding labiovelar stop: see below.
This was originally proposed by Vaillant, and has been reinforced by Kortlandt (2007 = 2009: 39â41).
Cf. Beekes (2011: 151). A similar rule has been proposed for Latin by Meiser (1998: 63â64): *rÌ¥ would have yielded ur after labiovelars and after u̯â. This rule is not widely accepted, however: cf. Zair (2017) for criticism. Frotscher (2012) has proposed that word-final *rÌ¥ in Proto-Italic developed to âur after labials (Lat. femur âthighâ, iecur âliverâ), but âer elsewhere (Lat. iter âroadâ, Å«ber âudderâ < *ouþrÌ¥).
Cf. LIV2 (s.v. ?2.*(s)kelâ).
Meier-Brügger (1990) proposed that ÎºÏ Î»Î»á½¹Ï âcrooked, club-footedâ reflects *kwlÌ¥nóâ, with the root *kwelâ âturnâ. This is contradicted by the reflex of e.g. *gwlÌ¥nâ in Î²á½±Î»Î»Ï (on which see section 9.5.1). I therefore hesitate to accept this etymology. To compare the pair ÎºÏ Î»Î»á½¹Ï / Î²á½±Î»Î»Ï with Î³Ï Î½á½µ beside Boeotian βανά, as done by Meier-Brügger (1990: 31 with n. 7), is not to the point: in the latter case we are clearly dealing with dialectally different treatments of the same word.
Compare also Îºá½»ÎºÎ»Î¿Ï which may reflect *kwokwloâ, *kwÉkwloâ or even *kwekwloâ. Cases of Cowgillâs Law (e.g. νύξ, á½Î½Ï ξ) are perhaps not directly comparable because their full vowel was colored by a contiguous labiovelar under more specific circumstances.
Incidentally, this proves that Lindemanâs Law in Greek is an inner-Greek affair, rather than an inheritance from PIE.
Chantraine (DELG s.v.) also analyzes ÎºÏ Ïâ as the u-colored reflex of a zero-grade root, but he does not explain how the vowel originated. There is no need to ascribe the u-vocalism to the allegedly expressive nature of this word, as per Bernabé (1977: 281).
Compare also Av. karÉna- âdeafâ, Latv. kurÌns âid.â (perhaps via *âwith defect earsâ), and with a different suffix Ved. krÌ¥dhú- âmaimedâ.
Various Balto-Slavic words denoting corporeal defects are derived from the PIE root *kwerâ. The consistent reflex (*)ur of the vocalized zero grade in these words confirms the idea of a conditioned rounded outcome of the syllabic sonorants in Balto-Slavic after labiovelars (Kortlandt 2007).
EM s.v. ÏαÏÎºá½±Î¶Ï (Kallierges 708): á¼ÏÏ Î¼á½½ÏεÏον δὲ Î»á½³Î³Î¿Ï Ïιν οἱ Îἰολεá¿Ï Ïá½»Ïκα, ÏαÏá½° Ïὸ á¼ÏοÏá½»ÏεÏθαι Ïὸ δέÏμα á¼Ïâ αá½Ïá¿ÏÎ Ïá½°Ï Î³á½°Ï Ïá½±ÏÎºÎ±Ï Ïá½»ÏÎºÎ±Ï Î¿á¼± ÎÏÏιεá¿Ï Î»á½³Î³Î¿Ï ÏιΠÏαÏá½° Ïὸ Ïá½»ÏÏ Ïá½»ÏξÏ, ÏὺÏξ καὶ Ïá½±Ïξ.
Schindlerâs argument for reconstructing an ablauting paradigm *tu̯orkâ beside *tu̯rÌ¥kâ is that the expected outcome of *tu̯rÌ¥kâ would be *ÏÏακâ on account of ÏÏá½±Ïεζα < *tu̯rÌ¥pedi̯a. This cannot be upheld: see sections 2.5 and 2.6 for a full discussion of ÏÏá½±Ïεζα and related issues.
In fact, this *tu̯ór-toâ could be identical to the pre-form required for the Avestan adj. θβÄÅ¡Ì£a- âquickâ (on this word, cf. EWAia s.v. TVAR).
Moreover, evidence for a simplification *tu̯rÌ¥ > *trÌ¥ under certain conditions must be taken into account: cf. section 2.7 on *kwetrÌ¥â âfourâ and *kwétrÌ¥toâ âfourthâ.
For instance in Risch (1955), Bader (1969).
Cf. Thompson (2010: 191), with a discussion of the most important Mycenaean material, citing a-ki-ti-to /aktiton/ âuncultivatedâ < *nÌ¥â and dat. pl. te-ka-ta-si /tektasi/ âbuildersâ < *tektnÌ¥-si.
Cf. Wathelet (1970: 175), who also remarks that an earlier vocalization of the syllabic nasals (as compared with the syllabic liquids) is paralleled in Indo-Iranian.
Thus, for Mycenaean, Lejeune (1972: 198), Leukart (1994: 110), Sihler (1995: 98). These three authors assume that only preceding labials could color the outcome, but Vine (1998: 35) argues that both preceding and following labials could cause this effect. He adduces Myc. o-wi-de-ta-i and o-mi-ri-jo-i as possible examples, forms which he interprets as *nÌ¥-u̯idetÄhi âto the invisible [deities]â and *nÌ¥-mrioihi âto the immortalsâ. More recently, Hajnal-Risch (2006: 212â213) and Thompson (2010: 191â192) argued in favor of the labial conditioning. See Thompson (1996â1997: 316â320) for an overview of the potential Mycenaean evidence for *NÌ¥.
It has been repeatedly observed (e.g. Ruijgh 1961, Wathelet 1970) that much of the alleged evidence for *nÌ¥, *mÌ¥ > o is found in the numerals. However, the analogical spread of o-vocalism through the numerals in certain dialects can in my view only be explained if there was a sufficient basis for the leveling. I agree with Thompson (1996â1997: 319) that it is difficult to explain Myc. e-ne-wo by analogy.
A less secure example is do-po-ta âlordâ < *dmÌ¥-pot-Äâ beside da-ko-ro < *dmÌ¥-koroâ âtemple servantâ, both from Pylos. Myc. do-po-ta is the recipient of an offering, and therefore most probably a theonym (cf. δεÏÏá½¹ÏηÏ). The reconstruction *dmÌ¥-pot-Äâ, however, is by no means certain: an o-grade *domâ cannot be excluded. Myc. da-ko-ro is an occupational term, and usually compared with class. ζάκοÏÎ¿Ï âtemple servantâ.
In οἰεÏέαÏ, οἰâ spells (metrically lengthened) /Å/ before a following ε. The third form with âcopulativeâ á½â in Homer is acc. pl. á½ÏÏιÏÎ±Ï (Il. 2.765). It could be argued that its á½â was taken over from οἰεÏέαÏ, which directly follows it in the same line. Homeric á½ÏαÏÏÎ¿Ï is clearly an archaic form, because it is attested twice in the verse end καÏίγνηÏÎ¿Ï ÎºÎ±á½¶ á½ÏαÏÏÎ¿Ï (Il. 12.371; acc. sg. Il. 11.257). Two other attestations of copulative á½â are found in Hsch.: á½Î¶Ï γεÏÎ á½Î¼á½¹Î¶Ï Î³ÎµÏ and á½Î³á½±ÏÏÏÏÎ á½Î¼Î¿Î³á½±ÏÏÏÏ.
Hajnal (in Hajnal-Risch 2006: 212â¯ff.) summarizes various proposals. With Thompson (1996â1997 and 2002â2003), I am pessimistic about the possibility to distinguish âmycénien normalâ from âmycénien spécialâ.
E.g. Ruijgh (1961: 205; 1967: 100â101), followed by Wathelet (1970: 173â175).
According to Ruijgh (e.g. 1961: 203), the form e-ka-ma-pi shows that the labial environment cannot be responsible for the rounded outcome.
Ruijgh (1961; 1985: 153â¯ff.).
However, an alternative is that the literary Lesbian forms in âÎ±Ï are epicisms or borrowings from Ionic (see section 3.3.3) or that they display the Pan-Greek regular reflex of word-final *ârÌ¥.
And also to the word for âearâ (cf. Myc. a-no-wo-to), which belongs to a different type of heteroclitic.
Hinge (2007: 146â147) has proposed that the o-colored reflex was conditioned by a following sibilant in εἴκοÏι (with assibilation, as opposed to Dor. ἴκαÏι) and a few other cases. Although this thesis has not found wide acceptance (no doubt because several of the etymologies provided by Hinge are not compelling), I agree with him that εἴκοÏι can hardly be an analogical form.
See sections 2.6, 2.7 and 3.3.1. Ruijgh (1961) also explained the numerals by analogical developments, but without invoking a conditioned change *nÌ¥, *mÌ¥ > o. In his view, Arc. δεκο, Thess. Lesb. δεκοÏÎ¿Ï and forms of ânineâ (cf. Lesb. ενοÏοÏ) would have acquired their final âο from á½ÎºÏο (attested as such in Boeot. and Lesb.), which would itself have developed from á½ÎºÏá½½ under influence of δύο beside δύÏ. Subsequently, the final âο would have spread to ânineâ and âtenâ. This scenario is accepted by Haug (2002: 51). However, even if influence of âtwoâ on âeightâ is accepted, it is remarkable that in a sequence á¼ÏÏá½±âá½ÎºÏá½¹âá¼Î½Î½á½³Î±âδέκα, it was the final vowel of á½ÎºÏο that prevailed. In my view, it would be much easier to explain the cases of o-vocalism in the numerals if âeightâ was assisted by ânineâ (Myc. e-ne-wo is the regular reflex of *eneu̯nÌ¥ due to the preceding labial consonant; similar for Lesb. ενοÏοÏ, and see now also Arc. ενÏοÏοÏ, Carbon-Clackson 2016). From ânineâ and perhaps also âeightâ, the final âo could spread to âtenâ. The forms δεκο, δεκοâ, δεκοÏÎ¿Ï (the latter attested in Thess. and Lesb., and possibly in Myc. de-ko-to PN) could then easily induce further analogical changes, such as Arc. ÏεμÏοÏÎ¿Ï for ÏέμÏÏοÏ.
For the Celtic evidence, see section 9.4.
The etymology of this example is misunderstood by KuryÅowicz, because á¼Î¼Î²ÏαÏαι is derived from the root of μείÏομαι âto receive as a shareâ, but νημεÏÏá½µÏ from that of á¼Î¼Î±ÏÏεá¿Î½ âto missâ. Moreover, á¼Î¼Î²ÏαÏαι is not an Ionic-Attic form (it is ascribed to the Syracusan mimographer Sophron); see section 3.2.2.
For Osthoffâs explanation, see section 1.4.5 below.
âIn manchen Fällen konnte das Nebeneinander von TRaT u. TaRT auch alte TReT-Wurzeln in Anspruch nehmen, so κÏεÏâ (κÏείÏÏÏν, äol. κÏá½³ÏοÏ), wo neben κÏá½±ÏοÏ, κÏαÏεÏá½¹Ï, κÏá½±ÏιÏÏοÏ, κÏαÏá½»Î½Ï auch die entsprechenden Formen mit Î±Ï auftreten. Ferner findet sich neben ÏÏá½³ÏÏ âgerinnen lassenâ ÏαÏÏá½»Ï âdick, geronnenâ, ÏαÏÏειαί, aber das Jonische hat auch ÏÏαÏεÏá½µ (γá¿) âfeste Erde, Festlandâ.â (KuryÅowicz 1968: 247).
For instance, Güntert mentions γÏá½±ÏÏ âto writeâ; this is indeed related to G. kerben âto carveâ, but the Greek dialects show evidence for an o-grade γÏοÏâ, the vowel slot of which might be older than that of the Germanic word. Other examples adduced by Güntert include βÏάκανα (n. pl.) âwild vegetablesâ (Pherecr. apud Ath. Deipn. 7.102; lexicographers) beside OHG moraha âcarrotsâ (these words are clearly borrowings; Greek βÏάκανα is closest to Ru. borkanâ âwild carrotâ: see Kroonen, EDPG s.v. *murhÅnâ with references); á¿¥á½±Î´Î±Î¼Î½Î¿Ï âbranchâ (LXX) has a variant á½Ïá½¹Î´Î±Î¼Î½Î¿Ï (Thphr.+); ῥάÏÏÏ âto sewâ has no clear etymology (it is not related to Lith. verpiù âI spinâ in view of forms like Myc. ra-pte-re); the root vowel of δÏá½±ÏÏομαι âto grasp with the handâ may be the reflex of a syllabic nasal (section 9.2.1); and the comparison between ÏÏá½±Î¼Î¹Ï âperineumâ and PGmc. *þarmaâ âintestineâ is a mere root etymology, cf. the judgement of Kroonen, EDPG q.v. (âpotentially related ⦠No further etymologyâ). The dubious status of a number of these etymologies is discussed in more detail elsewhere in this book.
Rix (1992: 65) only mentions the possibility of analogical influence of the full grade. Sihler (1995: 92) explicitly admits that the problem has not yet been solved, and gives a fairly neutral characterization: âThe conditions governing the appearance in Greek of αλ and Î±Ï vs. λα and Ïα have not been determined. In some words the difference is dialectal, but not in most.â The arguments and conclusions of OâNeil (1971) are so manifestly misguided that they require no extensive discussion. Idiosyncratic ideas about the coloring of the anaptyctic vowel are found in Wyatt (1971) and Bernabé (1977), but these authors do not deal with the place of the anaptyctic vowel, the issue with which we are especially concerned here.
Chantraine (1958: 23).
Kretschmer (1892), Schwyzer (1939: 342), Klingenschmitt (1974: 275), Hajnal-Risch (2006: 102â103; 202â205).
Hirt (1901: 232â238), Lejeune (1972: 196â197), Risch (1979a: 98â99), Thompson (2002â2003: 355â362), Hajnal-Risch (2006, l.c.).
KuryÅowicz (1956: 174â187; 1968: 243â247), GarcÃa Ramón (1985), Hajnal (1997: 145â150).
Osthoff (1879: 144â145), Hoenigswald (1953; 1968; 1988), Lubotsky (1994: 97).
Chantraine (1958: 23â24) gives the following discussion: âà lââ¯attique καÏδία «â¯cÅurâ¯Â» répond généralement lââ¯homérique κÏαδίη: καÏδίη nââ¯est possible quââ¯au nominatif et au datif singulier devant un mot à initiale vocalique, lââ¯hiatus abrégeant la longue finale (â¦); lââ¯ionien-attique a employé concurremment θάÏÏÎ¿Ï et θÏá½±ÏÎ¿Ï Â«â¯audaceâ¯Â»; le dialecte homérique a normalement θάÏÏÎ¿Ï (12 exemples), et une seule fois θÏá½±ÏÎ¿Ï (â¦); il existe une répartition entre καÏÏεÏá½¹Ï Â«â¯fortâ¯Â» qui est attique et κÏαÏεÏá½¹Ï, cette seconde forme étant employée lorsque la syllabe finale est longue; suivant les besoins du vers Homère emploie soit Ïá½³ÏαÏÏÎ¿Ï Â«â¯quatrième» qui est attique soit Ïá½³ÏÏαÏÎ¿Ï (â¦) qui, avec une finale brève, fournit une dactyle; enfin à βÏÎ±Î´á½»Ï répond un superlatif βάÏδιÏÏÎ¿Ï (â¦); βÏάδιÏÏÎ¿Ï ne pouvait entrer à aucune place du vers homérique.â
For such a resignation, see Goldstein (2013): âThe alternation between ra and ar or la and al may have been to some extent conditioned by speech tempo and register. As such, the precise conditions of their distributions may be unrecoverable.â
For instance, in Herodotus all instances of κÏαÏεÏá½¹Ï appear in oracles or otherwise clear epic reminiscences; the normal form is καÏÏεÏá½¹Ï.
According to Frotscher (2012) the reflex of word-final *ârÌ¥ in Indo-Aryan was also dependent on the accent: in his view, unaccented *ârÌ¥ yielded âar, as against accented *ârÌ¥Ì > âúr.
Cf. Haug (2002: 52) and section 1.2.3 with n. 55 above.
Hirt (1901: 238) further believes that Homeric âαÏâ may be due to metrical constraints: in pairs like κÏαÏâ / καÏÏâ, θÏαÏâ / θαÏÏâ, á¼ÏαÏÏιÏá½¹Ï / á¼ÏÏαÏιÏá½¹Ï â[liegt] bei Homer kein beliebiger Wechsel von Ïα und Î±Ï vor, sondern Î±Ï findet sich da, wo wir metrische Dehnung erwarten sollten.â But: âDass damit freilich noch nicht alle Î±Ï des Griechischen beseitigt sind, sehe ich wohl, indess glaube ich doch annehmen zu können, dass Ïα der alleinige Vertreter von rÌ¥ istâ. On an earlier occasion, Hirt had remarked: âDie Hauptargumente für unsere Ansicht werden bleiben: der Lok. Plur. ÏαÏÏá½±Ïι, und á½ÏόδÏα(κ) zu δέÏκομαι.â (Hirt 1897: 158).
The problematic instances of âαÏâ have often been tucked away in previous treatments. A good example is the discussion by Güntert (1916: 69â74). On the one hand, he accepts Kretschmerâs accent-conditioned development, but in addition he claims that Hirtâs discussion (which starts from the assumption of liquid metathesis) has shown âdass es kaum noch erwartet werden kann, in jedem Einzelfall die Verteilung von Î±Ï und Ïα zu erklären.â In this way, the hypothesis is protected against undesired falsificationâclearly an ad hoc strategy.
Schrijver has proposed a conditioning by adjacent pure velars (1991: 425â435), or an early phonologization of epenthetic vowels as /a/ before three consonants (1991: 488â498), but these suggestions do not seem to have met with general assent.
KuryÅowiczâs claim that a secondary zero grade /CRaC/ was also formed to roots of the structure *CRHC by analogy with roots of the type /CaC/ < *CHC is rightly problematized by Hajnal (1997: 146), who notes that Greek has a triple reflex of *CHC. On this basis, Hajnal (1997: 146â149) also criticizes KuryÅowiczâs claims concerning an Indo-European origin of the secondary zero grade. Nevertheless, Hajnal retains the concept of secondary zero grades in order to account for âaRâ as having a later, inner-Greek origin.
Hackstein (2002: 205â238) has argued for the spread of an analogical full grade á¾± beside á¾° in Greek, e.g. pf. Ïέθηλα âto be abundantâ beside θάλλÏ, ÏÎµÎ¸Î±Î»Ï á¿Î± (quasi PIE *dhe-dhlÌ¥h1-us-ih2), and pf. μέμηλα, ptc. μεμᾱλόÏâ (Pi.) beside Î¼á½³Î»Ï âto concernâ, μελέÏη âcareâ (root *melh1â).
DELG defines the original meaning as follows: âÏκάνδαλον consistait en une barre de bois plus ou moins longue qui constituait, soit une partie dââ¯un piège, soit la perche dââ¯un acrobat.â
The appurtenance of Celtic forms (OW serr âsickleâ, MIr. serr f. Ä-stem âid.â) is more uncertain. See section 9.6.1.
For a more detailed criticism of GarcÃa Ramónâs interpretation of καÏÏá½¹Ï and a number of Mycenaean forms, see section 2.2.
Cf. also Hoenigswald (1953; 1988).
The same is true of Grammont (1948: 285â286), who also ascribed the fluctuation between Î±Ï and Ïα as reflexes of *rÌ¥ to the rhythmical structure of the preceding syllable. Hoenigswald (1953: 289â290) claims that he found a confirmation of his idea in the concrete distribution of the particles á¼Ï and ῥα (á¿¥â) in Homer. In his view, these forms represent different vocalizations of *rÌ¥ depending on the weight of the preceding syllable. Originally, á¼Ï would be found after closed syllables with a bimoraic nucleus (e.g. Ïὴν á¼Ï), while ῥα would be used after long vowels, diphthongs and closed syllables with a short vowel (e.g. Ïῠῥα, ἦ ῥα, Ïόν ῥα, but rarely Ïήν ῥα). Hoenigswald points out that the type Ïήν ῥα occurs only 8 times on 91 occurrences of unelided ῥα in the first twelve books of the Iliad. However, it is not clear how significant this distribution is. In any case, even if this distribution were significant, it does not follow that ῥα (á¿¥â) reflects a pre-form *rÌ¥, as the particle was clearly utilized widely in Homeric Greek to make position length or to gain a syllable; this fact by itself explains why ῥα normally does not follow syllables that are long by nature. Moreover, I doubt whether the pre-form of á¼Ï and ῥα was *rÌ¥ (perhaps it was rather *hrÌ¥, see section 1.2.3).
As for ÏÏÏÎ·Î³á½·Ï âseal; brandmarkâ, it is unclear whether this reflects a full grade root *sbhreh2gâ or a zero grade *sbhrh2gâ. For the etymology connecting this word with ÏÏαÏαγέομαι âto hissâ, see Tichy (1983: 178â180) and Rico (2002).
As far as the examples adduced by Lubotsky are concerned, it is possible that ÏÏαλâ should be reconstructed as *sgwhh2elâ (cf. LIV2 s.v. *(s)gwhh2elâ and EDL s.v. fallÅ). The vowel slot of ÏθάÏμα can easily be secondary, cf. the full grade seen in ÏθείÏÏ âto destroyâ and its pf. âá½³ÏθοÏα, cf. also pf. mid. âá½³ÏθαÏμαι, aor. ÏθαÏá¿Î½Î±Î¹ âto perishâ. The pre-form of ÏÎºÎ±Î»Î¼á½¹Ï probably did not contain *lÌ¥ (see section 10.1.10); the forms ÏÏá½±Ïξαν, ÏÏá½±Ïγανα lack a clear etymology, meaning that they could owe their α-vocalism to the fact that they were borrowed.
In the words of Chantraine (1958: 111), âtoute la morphologie est commandée par des préoccupations métriques et nous aurons à chaque instant à faire appel à cette considérationâ. For artificial word-formation in Epic Greek, see Meister (1921), the papers collected in Witte (1972), and the recent overview by Hackstein (2010).
As we have seen in section 1.1.1, some scholars even claim (though without good grounds) that *r̥ was vocalized in all Greek dialects as early as the middle of the second millennium BCE. However, the interpretation of the Mycenaean evidence for the reflexes of *r̥ is not clear-cut: as I argue in chapter 2, a retention of r̥ in the Linear B tablets is not to be excluded.
For obvious reasons, I do not wish to take a strong position in the debate about the date of the Iliad and Odyssey, and about the genesis of the text. Nevertheless, if we assume that the largest part of both epics was composed somewhere between 750 and 650â¯BCE (the Iliad earlier than the Odyssey), this will in my view not be far from the truth.
This is not the place to go into further details. Contrary to the views expressed in Van Beek 2013, I am no longer strongly opposed to assuming the presence of an Aeolic element in Epic Greek. My current impression is that Epic Greek does have a number of old mainland Aeolic features, but these entered the tradition at a relatively early stage, in the Mycenaean or sub-Mycenaean period. Arguments favoring this third alternative (a poetic koinè stemming from the Mycenaean period) have been adduced, among other scholars, by Hooker (1977) and above all by Hoekstra (1981).
Even in the case of Attic, the best-known Ancient Greek dialect, it is notoriously difficult to pin down exactly what the ârealâ spoken language looked like (the style of all classical authors is elevated to a certain degree). For present purposes, however, it suffices to observe that the language of epic poetry has various characteristics (lexical, morphological, syntactic and stylistic) that are absent from texts belonging to other registers.
See Witte 1972; Forssman 1991; Janko 1994: 8â19; Hackstein 2010, among others.
I consider the cosmogonic aspects of Hesiodâs Theogony and the didactic parts of the Works and Days to be secondary genres with respect to heroic poetry and catalogues, although the former two genres may of course have some antiquity beyond Hesiod.
Whether the composer(s) of Iliad and Odyssey made use of writing or not is irrelevant here.
See section 6.8.7 for reasons why Î»Î±á½¹Ï is probably of Mycenaean origin.
See Hackstein (2010) for a convenient overview of artificial Homeric features.
Cf. Parry (1971: 331) and section 6.7.
One might object to this that the oldest hexameter inscriptions from non-Ionic-speaking regions usually contain non-Ionic phonology. For instance, the Mantiklos inscription (CEG I 326, Boeotia, 700â675â¯BCE) contains the forms ÏαÏá½·ÏεÏÏαν and Ïεκαβόλοι. However, this point is not probative for the issue under discussion. First, most of the phonological features (á¾± for η, retained Ï) are archaisms with respect to the corresponding Homeric features (in principle this may also hold for â¨ÏÏâ© against Homeric â¨ÏÏâ©, as we are not informed about the exact phonetic value of the spelling â¨ÏÏâ© in Boeotian at this early stage). Secondly, the tendency towards a more local orientation in archaic hexameter inscriptions (which undeniably exists on a morphological level: cf. Ïá½» for Ïá½» and the imperative δίδοι in the Mantiklos inscription) might well be a relatively recent development of the 8th and 7th centuries. Third, it is plausible that dedications and funerary epigrams, embedded as they were in a specific local context (and necessarily written in a local script), were more prone to absorb local features than poetry performed at festivals. Thus, nothing forces us to assume that poets automatically applied the phonology of their spoken dialect when performing in hexameter verse.
For compelling points of criticism regarding our ability to reconstruct a proto-hexameter, see Hoekstra (1981: 33â53).
E.g. Haug 2002; Hackstein 2002: 8â9; and extensively Hajnal 2003: 63â100.
As Barnes (2011: 9â10) remarks, âA problem with Tichyâs approach to these scansions has always been the implausibility of a scenario whereby not a single example of the phenomenon goes back to a form that would never have scanned properly.â For a similar criticism, see West (2011).
The forms and numbers are á¼Î½á½³Ïα 4â¯Ã, á¼Î½á½³ÏÎ¿Ï 19â¯Ã, á¼Î½á½³Ïι 9â¯Ã, pl. á¼Î½á½³ÏÎµÏ 41â¯Ã, á¼Î½á½³ÏÎ±Ï 17â¯Ã, and du. á¼Î½á½³Ïε 5â¯Ã.
This idea is repeated uncritically in Hajnal 2003: 78 n. 127.
Cf. á½Î½ ÎÏιάÏεÏν ÎºÎ±Î»á½³Î¿Ï Ïι θεοί, á¼Î½Î´ÏÎµÏ Î´á½³ Ïε ÏάνÏÎµÏ / ÎἰγαίÏνâ Il. 1.403â404, á½Î½ Îάνθον ÎºÎ±Î»á½³Î¿Ï Ïι θεοί, á¼Î½Î´ÏÎµÏ Î´á½² ΣκάμανδÏον Il. 20.74, Îºá½»Î½ÎµÏ Ïâ á¼Î½Î´Ïá½³Ï Ïε νομá¿ÎµÏ Il. 17.65, Îºá½»Î½Î±Ï Ïâ á¼Î½Î´ÏÎ±Ï Ïâ á¼ÏεθίζÏν Il. 17.658, á¼Îº á¿¥â á¼ÏÎ±Î¼á½·Î½Î¸Î¿Ï Î²á½°Ï á¼Î½Î´ÏÎ±Ï Î¼á½³Ïα οἰνοÏοÏá¿ÏÎ±Ï Od. 8.456. Since the thesis placement of á¼Î½Î´Ïâ does not occur very often, the restructuring of the paradigm may have taken place at a relatively early stage. Note that vowel-initial anapestic *á¼Î½á½³ÏÎµÏ was somewhat awkward to use, as it could not make position length. This may have helped to maintain the popularity of metrically lengthened dactylic á¼Î½á½³ÏεÏ.
The so-called hysterokinetic paradigm. In Greek, cf. also ÏαÏá½³Ïα, ÏαÏá½³ÏεÏ, ÏαÏá½³ÏÎ±Ï against ÏαÏÏá½¹Ï, ÏαÏÏá½·; and cf. Vedic acc. sg. pitáram, nom. pl. pitáraḥ, náraḥ.
Note, in passing, that assuming an earlier verse with *νá¿Î±Ï á¼Î¼ÏιελίÏÏÎ±Ï (allegedly with a trochaic fourth foot) does not take into account that the final syllable of the word preceding attested Î½á½³Î±Ï á¼Î¼ÏιελίÏÏÎ±Ï occupies the longum of the fourth foot, e.g. λιÏὼν Î½á½³Î±Ï á¼Î¼ÏιελίÏÏÎ±Ï Il. 17.612, á¼Ïον Î½á½³Î±Ï á¼Î¼ÏιελίÏÏÎ±Ï Od. 10.91.
The real problem, the high frequency of the gen. pl. form νεῶν as opposed to νηῶν, is not even mentioned by Berg and Haug, presumably because it is inconvenient for their thesis. Concerning this issue, Hoekstra (1965: 124â130) has argued that the shortened form νεῶν is in many cases due to modification of an older prototype with νηῶν.
Ideas that the hexameter is of very recent origin (Berg and Haug 2000), or even that it was coined by Homer (Tichy 2010), are devoid of all realism.
Niels Schoubben has recently elaborated this idea in an as yet unpublished Ghent MA-thesis written under the supervision of Mark Janse.
Cf. various remarks in this sense in Berg and Haug (2000), e.g. on pp. 9â10.