CC BY
36
The metrical evidence for pre-Mycenaean hexameter epic reconsidered1
Резюме. В статье оспаривается широко распространенная точка зрения, согласно которой в тексте Гомера формы ауддотцта, avfipsiyovт^, а^(Здота%о^5У, авдоту следует прочитывать с вокальным /г/. В свете того, что данный звук был утрачен уже в микенском греческом, подобный просодический архаизм указывал бы на сохранение у Гомера языковых элементов домикенской эпохи. В статье обсуждаются два альтернативных объяснения перечисленных форм (опущение назального и применение гипотезы Н. Берга об эолийском праразмере как основе гекзаметра). Ввиду несостоятельности уже предложенных объяснений, предпринимается попытка показать, что за этой просодической аномалией стоит редуцированная артикуляция эпентетических /d/ и /b/.
Ключевые слова: метрика; история гекзаметра; происхождение
гомеровского эпоса; историческая фонология; эпентеза
In the last couple of decades in the Anglo-American scholarship there emerged an apparent consensus, according to which dactylic hexameter and (one infers) narrative epic composed in this meter already existed in the pre-Mycenaean period. The authors of the recent relevant entries in the 1997 New Companion to Homer and the Neue Pauly, Geoffrey Horrocks and Joachim Latacz, consider the case proven. The linguistic evidence marshaled in support of this view is, however, less than secure. In particular, I will offer an alternative explanation of metrical anomalies in the Iliad that are currently widely regarded as traces of the vocalic /г/ - a phoneme that has been lost in all recorded dialects of Greek, including the Mycenaean.
The argument for dactylic hexameter in the pre-Mycenaean period rests on two pieces of linguistic evidence. The first is not difficult to lay aside. It concerns tmesis in Homer, that is, the use of preverbs (prefixes and adverbs) separately from the verbs; the preverb construction is not found in Mycenaean documents, which,
1 I am grateful to Lev Blumenfeld and David Goldstein for their comments on a draft of this talk and invaluable advice on matters of phonology, to Nikolaos Papazarkadas for his help with epigraphic material, as well as to Anna Morpurgo Davies who discussed the avfiQoryra problem with me at an early stage of my work on this topic.
however, does not entail that it was not available to speakers of Mycenaean Greek in other linguistic registers. As Dag Haug (2002: 42-44) and Ivo Hajnal (2004) have demonstrated, the texts in Linear B are hardly representative not only from the stylistic viewpoint, but also with regard to the pragmatic organization of discourse. The unmarked word order (object-preverb-verb), observed in Mycenaean documents, would give no occasion for the extraposition of the preverb, which, as Ivo Hajnal points out (2004: 175), can in fact be observed in pragmatically marked specimens of colloquial language in the classical period. Moreover, even if we regard the preverb construction as a syntactic archaism, it hardly proves the existence of verse epic, let alone of dactylic hexameter, in pre-Mycenaean times. Instead, tmesis was an age-old feature of Greek poetic language, which was inherited by the tradition culminating in Ionic epic - as well as by other poetic genres (cf. Morpurgo Davies 1985: 88). Crucially, Pindar’s usage does not differ from Homer’s in the frequency of preverb extraposition (Hummel 1993: 147). Thus, even though we can claim some continuity of the Homeric epic with poetic usage in the second millennium, we cannot link this continuity to a particular meter or genre (contra West 1988: 156, among many others).
The second argument is more complicated as well as more substantive. It has to do with the anomalous scansion of the word ауЬротута (‘- ^ - ^’ instead of ^ - ^’) and the related problem of scanning the hapax а(^)врота^о^гу, the formulaic phrase EvuaA/w ауЬргіфоутл, and the form авдоту. This problem has been remarked upon by Jakob Wackernagel in a footnote dating from 1909, where he observes the following feature of Homeric prosody:
“eine Silbe mit kurzem Vokal, dem ursprunglich Nasal + r folgte, bei Homer vor der Silbenfolge ^ - kurz gemessen werden konnte.”
In other words, whenever the word is otherwise metrically inadmissible in a hexameter line, the NPL [nasal + plosive + liquid] cluster may fail to make the preceding light syllable into a heavy one.
The solution of the “ауЬрот^та problem” that is most widely accepted today was foreseen by Wackernagel: it is to posit a vocalic /r/ and read *anMta,*anrtaxomen, etc. to obtain the right scansion (^ ^ - ^). Wackernagel himself mentions such a possibility only to reject it as too far-fetched: “Kaum kann fur die alteste Phase der epischen Sprache geradewegs noch sonantisches r vorausgesetzt werden” (1170 n.1) . Yet the vocalic /r/ hypothesis was destined for
2
West 1988: 156 cites Wackernagel when discussing the vocalic /r/ hypothesis, but neglects to clarify Wackernagel’s view of it.
a momentous afterlife in the Homeric criticism. After it was resurrected by Hugo Muhlestein (1958: 364-365), and found support of C. J. Ruijgh and M. L. West, the notion that Homeric prosody retains traces of the vocalic /r/ has become the dominant view3. Notably, at the time when Wackernagel was pondering this problem, positing a vocalic /r/ would not entail significant reconsideration of the prehistory of epic language. Since the decipherment of Mycenaean, however, it has become generally accepted that by the time of the Linear B tablets the vocalic /r/ in Greek already turned into a combination of a consonant /r/ and a supporting vowel4. Accordingly, if we accept the vocalic /r/ hypothesis today, we have to draw a much more far-reaching conclusion: the Iliad preserves a phoneme that has been removed from the language at least 700 years before the poem was committed to writing (assuming it was written down in the eighth century).
Let us begin by looking more closely at the evidence. M. L. West puts a lot of emphasis on the archaic nature of the most clearly formulaic context in which the suspected vocalic /r/ occurs: the four lines that describe Meriones as “equal to man-slaying Enualios” (2.651, 7.166, 8.264, 17.259): MyQiovyg т araXavrog EvuaXіш
avbpzicpovry (^ - ^ ^-------)5. The ending of this line is, strictly
speaking, metrical, but calls for a very unusual synecphonesis of the final vowel of EwaXiw and the opening vowel of avbpzicpovrj}: the two vowels are articulated as one (West 1982: 15). If rewritten with a vocalic /r/, we read *Enuwalioianrqwhontdi (West 1988: 156); here a
3 Muhlestein 1958: 364-365, Wathelet 1966, West 1988: 156, Ruijgh 1995: 88-91, Horrocks 1997; Watkins 1995: 499; Latacz 1998. Among those scholars who reject this hypothesis are: Tichy 1981, Berg and Haug 2000, Kullman 2001: 649, Hackstein 2002: 5-9, Haug 2002: 62-67, Barnes Forthc. (discussed in the appendix). Katz 2010: 365 (and p.c.) favors the hypothesis presented in Barnes Forthc.
4 Heubeck 1972: 74-9 accepts Muhlestein’s theory that the text of Homer contains traces of the vocalic /r/ only to use it as evidence for the preservation of this sound in Mycenaean, pointing to the absurdity of positing pre-Mycenaean survivals in Homer. For some scholars (e.g. Crespo 1977: 24), the possibility of postdating the loss of vocalic /r/ increased the plausibility of its survival in Homer, but Heubeck’s view has not won the acceptance of the Mycenologists (see recent overview in Haug 2002: 4962).
5 In our text of Homer, we also find an innovated, yet not metrically equivalent version of атаХаутод EwaXiw ауЬраїфоут^ in the formula ('Ектшд Пдіа^ііїпд) @дотоХоіуф іаод ’Kgn'i (Il. 11.295, 13.802; also Od. 8.115), with two lexical replacements: атаХаутод > 1<год and ауЬреїфоут'Г} > вдотоХоїуф. Tichy regards атаХаутод EwaXlw ауЬраїфбутгі as an adaptation of Sow атаХаутод ”Адцї (8.215, 13.295, 13.328, 13.528).
hiatus is avoided by a presumed preservation of the glide y in the dative singular (following the /o/; cf. Haug 2002: 63-64, Willi 2003: 225 n2) - a phenomenon unparalleled in our text of Homer. There are further reasons to doubt the archaism of ’EvvaXiw avdgs'iyovry (Tichy 1981: 39-40): if we posit regular historical derivation, the form avdQs'iy6vry goes back to an intermediate form *avdQoy6vry, of which it must represent a metrical normalization, as *avdQoy6vry is already unmetrical in the end of the line. The alternative view is to regard avdQs'i^6vry as modelled on the epithet of Hermes, Agysi^vrng (7x in the Iliad); notably, a dative form of this epithet occurs once, also at the end of the line: diaxr6Qw aQYs'i^6vry (2.103)6.
There is thus no compelling metrical reason to reconstruct a form with a vocalic /r/ in ’EvvaXiw avdQs'i^6vry, and there are serious impediments to treating it as an archaism. I have nevertheless dwelt on this phrase at some length because it is the only formula containing a presumed vocalic /r/ that can be plausibly projected back into the pre-Mycenaean period.
In the case of the form avdQoryra itself, whose prosodic shape (-^ -) makes the reconstruction of a form with a vocalic /r/ (anrtata) much more urgent, the formulaic context is uncertain. The phrase ov n6r^ov yo6w(ra Xmovv’ av^goryra xai yfiyv occurs twice in the Iliad (16.857, 22.363), and the word avdgoryra is attested in a different metrical position in narg6xXov no%wv av^Qoryra ts xai ^kvoc, yv (24.6). Should we wish to reconstruct a pre-Mycenaean protoformula equivalent of avdgoryra xai yfiyv, we run into the problem of Greek xai (at least in the function of word connector) being a late form; it is not attested in Mycenaean. The proponents of the vocalic /r/ hypothesis assume (e.g. West 1988: 156) that a reconstructed proto-formula would have been anrtat’ ide yegwan, with an elision of the final vowel of anrtata. Yet Ruijgh, who originally proposed this substitution, more recently withdrew it on semantic grounds (citing that ide in Cypriot means ‘alors’) and gave up on reconstructing the formula (1995: 90-91). The antiquity of the formation avdgoryra is also subject to doubt (Tichy 1981: 47-48). To the best of my knowledge, no alternative proto-form has been proposed.7 From the
6 Tichy also proposes that the proper scansion of EvuaAlw av$Qai<p6vTy (based on 8iaxT6QM agys'i^vTy) demands the consonantalization of ‘i’ in EvuaAlw, yielding e.nu.a.lio (^ - ^; /o/ short by correption); this scansion would altogether remove the metrical problem.
7 I was able to see a draft of the forthcoming paper by Timothy Barnes in December 2010, after the argument advanced here has been formulated. I discuss my disagreements with Barnes in an appendix that follows this paper.
viewpoint of context in which it occurs, antiquity of the formula can also be doubted: the notion of the soul “flying off” (ex gs^swv nra^svn), while cross-culturally widespread, appears to involve the Near Eastern concept of soul as breath (West 1997: 151, pace 2007: 490).
I would like to emphasize at this point a fact that has not been foregrounded in the recent work on the avdgoryra problem: the preservation of the ancient prosody of words that presupposes a vocalic /r/ is only possible in formulaic contexts; once a word becomes an independent entity with no formulaic associations or (in the very least) a set position in the line, its prosodic nature must conform to the current articulation, unless a case can be made for this word being preserved only in the poetic language.
If we keep this in mind, the remaining pieces of evidence that have been adduced in support of the vocalic /r/ hypothesis are revealed as being in fact detrimental to it. Of critical importance is the hapax form a^figora^o^sv (most MSS record the metrically normalized form afigora^o^sv) that occurs in Il. 10.65 (Agamemnon addressing Menelaos): aud-i ^svsiv, nwg afigora^o^sv aXXyXoiiv / sg%oiAsvw ‘Stay here, lest we miss each other somehow on our way’. The verb afigoraZw is related to Ionic a^agrslv and goes back to a root with a vocalic /r/ (< *h2mrt); another reflex of the Aeolic treatment of this sequence is attested 6 times in the Iliad alone in the aorist indicative forms (an)y^figor-. The suffix -a£- in afigora^o^sv is morphologically highly problematic; as Tichy (1981: 37-38) has shown, afigora^o^sv can only be a regular formation if the verb is a denominative from ^a^figoray- or *a^figorax-, neither of which is attested (regular presents in -aZs/o in Homer form aorists in -av(v)-). Whether or not we regard it as an artificial, metrically conditioned lengthened form, as Tichy suggests8, positing a 700-year prosodic survival for Il. 10.65 appears to be highly implausible, as the context is unique and indeed, from the viewpoint of a fairly unusual content, not likely to have ever been formulaic (whatever our definition of formula is).
Thus, rather than lending support to the vocalic /r/ hypothesis, the prosody of afigora^o^sv calls for a different approach to the prosody of NPL clusters. This by itself undermines the reconstruction of the vocalic /r/ in other, more formulaic contexts, as any alternative explanation would apply to all words with an imputed vocalic /r/.
The same reasoning applies to the fourth piece of evidence - the sole occurrence of the adjective afigorn in fem. nom. in Il. 14.78 (sig
o ksv sX&'fl / vv% afigorn, yv xai ry anovx^vrai noXk^oio / Tg&sg);
8 Tichy proposes that the form may be modeled on aXX’ tjtoi am wxti <pvXa^o^av y^aag avTovg (Il. 8.529).
the adjective is otherwise spelled a^fiqorog and scanned with a heavy first syllable. The morphology of afigorn is unexpected, as compounds with verbal adjectives in -to in Homer do not generally have a separate feminine form, except when an influence of the simplex form can be observed - which is not the case with figorog that lacks a feminine form (as pointed out by Tichy 1981: 34-37). In light of the fact that this usage is unique, the reconstruction of a traditional formula that could preserve an ancient prosody appears unwarranted.
The aberrant morphology of afigorn appears to be locally conditioned: it replaces a^figoain, which is the usual feminine form used with the noun vv£ in Homer9. Paradoxically, the poet appears to have produced a metri gratia replacement form that is highly problematic precisely from the metrical viewpoint. This consideration favors the view (maintained by Tichy 1981) that afigorn was formed by analogy with a^pifigorn, used in the Iliad with the genitive (2.389, 12.402, 20.281) and the accusative (11.32) of aanig, which appears to be a late formation (derived directly from figorog, rather than *mrtos). On this view, both of these forms are irrelevant to the prosody of NPL clusters. (On alternative explanations of the avdgoryra problem, discussed below, we do not need to posit late derivation for these forms, as they would fall under a phonologically-based metrical license.)
More generally, however, it should be emphasized that the vocalic /r/ hypothesis depends exclusively on words whose aberrant scansion is the condition of their entry into the epic diction. We never find words with a historical vocalic /r/ that are prosodically ambidextrous, so to speak, such as agva < *frna, in formulas that demand the scansion ^ ^ (for similar criticism, cf. Tichy 1981.54). Nor do we find formulas such as *a^figorog Ssog ^ ^ ^ ^, where
the quantity of the first syllable of a^figorog would be dictated not by the prosodic shape of the word, but by the formulaic environment.
These ex silentio arguments against the vocalic /r/ hypothesis are serious, but hardly fatal. Much more damaging is the sheer implausi-bility of a^figora^o^sv as a prosodic archaism and the difficulty of reconstructing a 700-year-old formula containing avdgoryra.
These weaknesses of the vocalic /r/ hypothesis demand that we consider alternative approaches to the prosody of NPL clusters. I would like to survey three such approaches, two of which have been
9 Accusative of vvxra appears in the phrase vvxra Bi’ a^figotr'inv in the first hemiepes in Il. 2.57, 10.41, 10.142, 24.363 (and in Od. 9.404, 15.8); a^$Qo<rin stands first in the line, qualifying the nominative in the preceding line, in Il. 18.268. a^figotr'in is also at the end of the line in
Od. 4.574, 7.283.
discussed in the scholarship, and one is presented here for the first time.
The first approach is based on the orthographic - both epigraphic and textual - evidence for nasal omission before stops in Archaic Greek. This evidence was generally held to provide a solution to the avdporyra problem before the (paradoxical) ascendancy of the vocalic /r/ hypothesis in the wake of the decipherment of Linear B. Thus, Wackernagel pointed to the Homeric MSS spelling afipora^o^sv where we would expect a^fipora^o^sv (1955: 1116, n. 1); similar MSS variation is attested for avdporyra, spelled as adporyra. In 1965, Joachim Latacz argued that AAPOTHTA was the original spelling in Homer’s text, pointing to the epigraphical evidence that before certain consonant clusters nasal sounds could be reduced (“als fluchtige Ubergangslaute behandelt und dementspre-chend im Schriftbild ausgelassen werden”; 1965.66). The same solution is assumed by Pierre Chantraine in Grammaire homerique (1958-1963: 1.109), who in relation to the avdporyra problem speaks of “la debilite de la nasale en grec.”10
Several criticisms can be made of this view. The spelling in the Homeric MSS may easily be explained away by metrical normalization (West 1982: 26). More broadly, the phonological nature of the reduction of the nasal is uncertain. The effect of denasalization before voiceless consonants is well-attested cross-linguistically; much less commonly, it spreads to voiced consonants (Pater 1999, Blust 2004: 101-36). Thus, in Modern Greek epempsa is pronounces as epepsa, but no nasal deletion occurs in the case of andras (Grimes et al. 2002). This reasoning does not rule out the possibility that nasal reduction was ubiquitous, and it just happens to assume metrical significance before voiced consonants in the Iliad.
Secondly, there are reasons to believe that the omission of the nasal in writing points to gemination, rather than nasal reduction, as suggested by Schwyzer (1939: 214); yet here the evidence is not conclusive, as Schwyzer admits.
Thirdly and, in my view, most importantly, if nasal reduction were available to the poet as a metrical license, one would expect it
10 For an earlier formulation of this point see Humbert 1938. Leslie Threatte believes that “the phenomenon [of the omission of nasals before stops in Attic inscriptions] is sufficiently well documented to justify the view that there existed a tendency to drop the nasal in this position, possibly with nasalization of the preceding vowel”; he further notes that “many of the examples involve a nasal before a consonant cluster” (1980: 485). It has, in fact, been argued that a similar orthographic practice of omitting nasals in Hittite has a phonetic basis (Justeson and Stephens 1981).
to be attested more broadly. It would indeed be a strange coincidence if it applied only in NPL clusters.
The second alternative explanation of the avdgoryra problem was presented by Eva Tichy in 1981; it relies on the metrical derivation of hexameter proposed by Nils Berg in 1978. On Tichy’s account, avdgoryra xai yfiyv is a pherecratean formula that predates the hexameter and is for that reason metrically ill-fitted for this meter11. Tichy’s argument has been accepted, for example, by Michael Meier-Brugger, who rules out a pre-Mycenaean background of the dactylic hexameter (1992: 1.93; similarly: Berg and Haug 2000, Hackstein 2002: 5-9, Haug 2002.). Apart from the fact that it presupposes a reconstruction of the proto-form of the hexameter that remains speculative, this explanation is problematic in two respects: (1) it separates the avdgoryra xai yfiyv formula from all other anomalous scansions of NPL clusters discussed above; yet Tichy fails to provide a convincing explanation for the scansion of afiQora&iAsv12; (2) Tichy has to fall back on the nasal reduction hypothesis to explain the usability of the formula in the hexameter at a stage that postdates its original employment (Tichy 1981: 46 n.41, 60-61).
I would like to propose a third explanation to the avdgoryra problem. As is well known, all attested Greek dialects (including Mycenaean) are intolerant of consonantal sequences [mr], [ml], and [nr]; in such sequences “the end of the nasal sound loses its nasality and is reduced to a plosive [b] or [d]” (Lejeune 1974: 154). These epenthetic consonants were very brief and often not marked in inscriptions (Lejeune 1974: 154; for a fuller discussion of the inscriptional evidence, see Appendix)13.
In archaic and classical prosody, some (but not all) plosive + liquid clusters, referred to as muta cum liquida, were treated as tautosyllabic; that is, they failed to make the preceding syllable heavy. In the middle of a word, in Homer such clusters were generally only treated as tautosyllabic when the word would
11 This does not mean that it has to be particularly old. For the suggestion that the hexameter emerged in the 8 th c., see Berg and Haug 2000.
12 Tichy 1981.39 tentatively assigns a quasi-formulaic status to the (hapax) phrase afigoTa&^sv aXXyXoiiv, metrically equivalent to (another hapax) avBQOTyTa ts xai ^avog yij'; the value of these “formulas” for Berg’s theory is in any case moot, as neither amounts to a pherecratean.
13 In fact, John Ohala (e.g. 1997) refers to epenthetic consonants as “emergent” consonants on the ground that all phonetic elements are already present (and so nothing is “additionally inserted”). Note also experimental evidence from Estonian showing that stops “were found to be particularly short after nasals in intervocalic clusters” (Devine and Stephens 1994: 68).
otherwise be banned from the hexameter line, as in the case of A^QoBtrn (full list in Tichy 1981: 30).
Since we do not have access to Greek poetry of the period when nasal + liquid combinations were still part of the language, we have no way of knowing whether they allowed for a similar license. Yet the hypothesis that consonantal clusters comprising [nasal + secondary plosive + liquid] could be treated as tautosyllabic cannot be ruled out. Homeric prosody is notably lenient in permitting otherwise inadmissable scansion of words that do not fit into the hexameter line; initial sequences [sk] and [zd] in Homer and Hesiod can be tautosyllabic. In some hexameters preserved in later sources [pt] is tautosyllabic; in classical texts [kt] and [mn] can be tautosyllabic (West 1982: 17-18).
A further, possibly decisive piece of evidence in favor of this hypothesis comes from the anomalous scansion of a^nXaxy^ara in Aesch. Eum. 934 (anap.), printed, uniquely, as anXaxy^ara metri gratia in modern editions. Tichy 1981: 64 discerns the significance of this scansion, which confirms that “a metrical shortening before NPL clusters, although it went against the rules of prosody, was not impossible in classical Attic” (“die metrische Kurzmessung vor NPL, obwohl gegen die prosodischen Regeln verstoBend, im klassischen Attischen phonetisch nicht volling unmoglich war”), but -implausibly, in my view - explains it by the influence of the scansion of a(^)fiQora^o^sv in Il. 10.65 - or of other (unattested) derivations of the root ambrot in epic. In fact, similar scansions are found -again, in anapaests - with a^nXax^v in Eur. Alc. 242 and Iph. Aul. 12414.
If this hypothesis is accepted, it would appear to contravene the universally accepted truth that the metrical evidence excludes the possibility of all three consonants being assigned to the onset” of the syllable (Devine and Stephens 1994: 42). Cross-linguistically, however, it is established that epenthetic stops are often not phonetically equivalent to inherited stops. For example, the epenthetic [t] in English dense is significantly shorter (and the nasal significantly longer) than the corresponding sounds in dents (Fourakis and Port 1986)15. Note that if a reduced plosive can be
14 For a discussion of these passages see W. Clemm, who argues for the weakness of the nasal sound in this position (“der labiale Nasal habe vor gewissen Consonanten zuweilen einen so schwachen Laut gehabt, dass er in den anapastischen Systemen von Euripides prosodisch ignoriert werden durfte”; 1877: 467).
15 Wetzels 1985 has argued that two distinct kinds of consonantal epenthesis are to be distinguished: (1) involving the initial sequence of two consonants, where the second one is liquid (e.g. grim[b]ly in some English dialects), and
posited for Ancient Greek NPL clusters, the syllable onset [nasal + liquid] would not contradict the sonority hierarchy.
The third explanation has the virtue of strictly defining the cases to which it applies (rather than having too broad a scope, like the nasal reduction hypothesis) and of treating together the prosodic anomalies usually classed under the vocalic /r/ hypothesis (in contrast to Tichy’s approach).
In conclusion, I would point to a parallel to the divergent metrical treatment of epenthetical sounds within a long poetic tradition. In Russian, sequences [br] [bl] [tr] [zn’] include a reduced vowel, as in dekabr’, korabl’, piupitr, zhizn’ (note that this vowel is not written, in conformance with an earlier pronunciation norm). In syllabotonic versification, this reduced vowel may gain prosodic significance, and many amateur (as well as, occasionally, professional) poets make use of this license. The non-observance of “the rule of the mute e” has often been cited by literary critics as a sign of dillentant versification (Maslov 2006: 70). In the Twenty Sonnets to Mary Stuart, Joseph Brodsky would scan Leandr as trisyllabic, when imitating a naive lyric persona. In this case, the use of the epenthetic consonant in everyday speech is resisted within the poetic tradition, which (for at least over a hundred years) adheres to an older pronuncation than that reflected in the orthographic convention 6. Moreover, owing to the reduced pronunciation of the epenthetic sound, the same word could appear in different prosodic variants within the corpus of the same poet, based on metrical convenience.
(2) involving the sequence of two consonants where the second one is not a liquid (e.g. prin[t]ce in American English). He further argued that whereas the epenthetic consonant in (1) is fully-fledged (and is motivated by syllable structure), in (2) it is a mere “contour”. Murray 1989 and Czaplicki 2010 argue against this distinction. In general, most of the work on epenthesis has focused on the phonetics of epenthetic stops of the prince variety (which is more important in Modern English), and no solid evidence seems to exist for the articulation of NPL clusters in English or any other language, comparable to that presented in Fourakis and Port 1986 for prince-type epenthesis.
1 Cf. Zygis 2003: 120: “As several studies show, the phonetic insertion can also be incorporated into underlying representation and orthography being a final product of an insertion process. Before the insertion enters the orthographic convention of a given language, it sporadically happens to appear in orthographical representation and is generally treated as a mistake. For example, the Polish word <sytuacja> is permanently written by children as <sytulacja>.”
Appendix on Barnes Forthc.:
Timothy Barnes (Forthc.), whose paper I saw after the argument here presented had been formulated, accepts Tichy’s explanation of all other anomalous NPL scansions, but argues that avdgoryra xai yfiyv was based on the reconstructed formula *a^fiQoryra xai yfiyv. On Barnes’s account, the substitution *a^fiqoTyTa > avdqoTyTa was driven by a semantic change (the original meaning he posits for the unattested noun *a^fiqoTyTa, on the strength of an Avestan parallel, is ‘non-dying’), yet had a metrically inadmissible result. (Barnes offers no explanation for the scansion of avdqoTyTa after the substitution took place.) This hypothesis contradicts the intrinsic logic of oral poetic tradition, which tends to preserve semantic and morphological archaisms even in spite of their prosodic difficulties; moreover, hexameter is known to develop toward more metrical strictness.
Similarly to the argument proposed here, Barnes accounts for the prosodic problem by the assumption of non-development of epenthetic [b] in [mr] sequences, citing inscriptional evidence in favor of this, as well as claiming a “complete absence of any instances of -nr-“ in inscriptions. In fact, sporadic omission of epenthetic [b] and [d] occurs in inscriptions in different periods: Avqo^a%z Attica, red-figure cup, early 5 c. BCE17; K[ki^avq\o<; Samothrace 323-316 BCE; AXs^avqs'iag Chios, imperial; avqi Lycaonia, no date; K^qoTy Zw<r[y\ Phrygia 175-200 CE; ya^qog Phrygia 2-3 c. CE18, which suggests persistent reduced articulation. The early inscriptional evidence for the omission of epenthetic [b] marshaled by Barnes could indicate the relative weakness of epenthetic [b], compared to [d]; this may indeed help explain the preponderance of prosodically aberrant forms with original [mr] in Homer.19 Yet the claim that “the development of the epenthetic consonant... may not have happened in (-)mr- sequences at all as of the fixation of the Homeric text” (MS p. 20) is problematic in light of
17
On this inscription, see Threatte 1980: 573, who notes: “The -d- in this cluster is of course itself an epenthetic consonant which developed to facilitate the pronunciation of [-nr-], and the original spellings in -vp- are also found in other parts of Greece; but whether these spellings are late survival of the original form or due to cluster simplification is uncertain.”
18 Results obtained through http://epigraphy.packhum.org/inscriptions/.
19 Note that the opposite tendency in the uneven development of epenthetic consonants can be observed in Old French dialects (Morin 1980: 209): “l’epenthese la plus generale est celle de [b] dans les suites [-mr-], cette derniere etant beaucoup moins marquee que l’ epenthese de [d] dans les suites [-nr-].”
the difference of scansion of amphibrotos ([br] always tautosyllabic) vs. ambrotos ([mbr] never tautosyllabic), which suggests that the former was derived directly from brotos. This difference is predicted on the hypothesis presented here (as muta cum liquida in Homer are more often tautosyllabic than NPL clusters), yet remains unexplained on Barnes’s account, which treats these forms as *amphimrotos and *amrotos. For Mycenaean, whereas we know that epenthetic [d] did develop in [nr] sequences, there is no evidence either for or against the development of [b]; the former, however, is a priori more likely (Lejeune 1974: 154).
Bibliography
Barnes - Barnes T. Forthc. Homeric ANAPOTHTA KAI HBHN // JHS. Berg 1978 - Berg N. Parergon metricum: der Ursprung des griechischen Hexameters // MSS 37. P. 11-36.
Berg, Haug 2000 - Berg N., Haug D. Innovation and Tradition in Homer -an Overlooked Piece of Evidence // SO. Vol. 75. P. 5-23.
Blust 2004 - Blust R. A. Austronesian Nasal Substitution: A Survey // Oceanic Linguistics. Vol. 43.1. P. 73-148.
Chantraine 1958-1963 - Chantraine P. Grammaire homerique. Vol. 1-2 Paris.
Clemm 1877 - Clemm W. AvBgor^g // RhM. Bd. 32. S. 463-74.
Crespo 1977 - Crespo E. Elementos antiguos y modernos en la prosodia homerica. [Suppl. to Minos, no. 7]. Salamanca.
Czaplicki 2010 - Czaplicki B. Emergent stops in English and Polish: Against Syllable-Based Accounts // Poznan Studies in Contemporary Linguistics. T. 46. P. 177-191.
Devine, Stephens 1994 - Devine A. M., Stephens L. D. The prosody of Greek speech. New York.
Fourakis, Port 1986 - Fourakis M., Port R. Stop epenthesis in English // Journal of Phonetics. Vol. 14. P. 197-221.
Grimes, Keil, Tippetts 2002 - Grimes S. M., Keil B. F., Tippetts K. NC-Phonology in Modern Greek // Ms. http://www.ldc.upenn.edu/ ~sgrimes/papers/NCPhonology.pdf Hackstein 2002 - Hackstein O. Die Sprachform der homerischen Epen: Faktoren morphologischer Variabilitat in literarischen Fruhformen: Tradition, Sprachwandel, Sprachliche Anachronismen. Serta Graeca 15. Wiesbaden.
Hajnal 2004 - Hajnal I. Die Tmesis bei Homer und auf den mykenischen Linear B-Tafeln: ein chronologisches Paradox? // J. H. W. Penney, ed. Indo-European Perspectives: Studies in Honour of Anna Morpurgo Davies. Oxford. P. 146-178.
Haug, Welo 2001 - Haug D., Welo E. The Proto-Hexameter Hypothesis: Perspectives for Further Research // SO. T. 76. P. 130-36.
Haug 2002 - Haug D. Les phases de revolution de la langue epique: trois etudes de linguistique homerique. Hypomnemata 142. Gottingen. Heubeck 1972 - Heubeck A. Syllabic r in Mycenaean Greek? // Minos 12 = Acta Mycenaea: Proceedings of the Fifth International Colloquium on
Mycenaean Studies. P. 55-79. Reprinted in Kleine Schriften zur griechischen Sprache und Literatur (Erlangen, 1984), 406-30.
Horrocks 1997 - Horrocks G. C. Homer’s dialect // I. Morris, B. Powell, eds. A New Companion to Homer. Leiden. P. 193-217.
Hoekstra 1981 - Hoekstra A. Epic verse before Homer. Amsterdam.
Humbert 1938 - Humbert J. Thucydide III, 104 et le vers 171 de l'Hymne homerique a Apollon // REG. T. 51. P. 275-81.
Hummel 1993 - Hummel P. La syntaxe de Pindare. Paris.
Justeson, Stephens 1981 - Justeson J. S., Stephens L. D. Nasal + Obstruent Clusters in Hittite // Journal of the American Oriental Society. Vol. 101.3. P. 367-70.
Katz 2010 - Katz J. Inherited Poetics // A Companion to the Ancient Greek Language / ed. E. J. Bakker. Chichester, West Sussex. P. 357-369.
Kullmann 2001 - Kullmann W. Rev. of: Homers Ilias, ed. J. Latacz and J. Latacz, Troia und Homer // Gnomon. Bd. 73. S. 648-63.
Latacz 1965 - Latacz J. ANAPOTHTA // Glotta. Bd. 43. S. 62-76.
Latacz 1998 - Latacz J. Epos. II: Klassische Antike // Der Neue Pauly. Vol. 4. Stuttgart. Col. 11-22.
Lejeune 1972 - Lejeune M. Phonetique historique du mycenien et du grec ancien. Paris.
Maslov 2006 - Maslov B. Traditsii literaturnogo diletantizma i esteticheskaia ideologiia romana ‘Dar’ // Imperiia N: Nabokov i nasledniki. Ed. Iu. Leving and E. Soshkin. Moscow. 37-73.
Meier-Brugger 1992 - Meier-Brugger M. Griechische Sprachwissenschaft. Bd. 1-2. Berlin; New York.
Morin 1980 - Morin Y.-C. Morphologisation de l’epenthese en ancien Frangais // Canadian Journal of Linguistics. Vol. 25. P. 204-25.
Morpurgo Davies 1985 - Morpurgo Davies A. Mycenaean and Greek Language // A. Morpurgo Davies, Y. Duhoux (eds.). Linear B: A 1984 Survey. Louvain-La-Neuve. P. 75-125.
Muhlestein 1958 - Muhlestein H. Interpretations de mots myceniens // Athenaeum. T. 36. P. 360-68.
Murray 1989 - Murray R. W. On Epenthesis // Folia Linguistica. T. 23. P. 293-316.
Ohala 1997 - Ohala J. J. Emergent stops // Proc. 4th Seoul International Conference on Linguistics [SICOL] 11-15 Aug 1997. P. 84-91. Online: http://linguistics.berkeley.edu/phonlab/users/ohala/index3.html
Pater 1999 - Pater J. Austronesian nasal substitution and other NC effects // The prosody-morphology interface / ed. H. van der Hulst et al. Cambridge. P. 310-43.
Ruijgh 1995 - Ruijgh C. J. D’ Homere aux origines proto-myceniennes de la tradition epique: analyse dialectologique du langage homerique, avec un excursus sur la creation de l’alphabet grec // J. P. Crielaard (ed.). Homeric Questions: Essays in Philology, Ancient History and Archaeology Amsterdam. P. 1-92.
Schwyzer 1939 - Schwyzer E. Griechische Grammatik. Bd. 1. Munchen.
Tichy 1981 - Tichy E. Hom. avBQoryra und die Vorgeschichte des daktylischen Hexameters // Glotta. Bd. 59. S. 28-67.
Threatte 1980 - Threatte L. The Grammar of Attic Inscriptions. Vol. 1: Phonology. Berlin.
Wackernagel 1955 - Wackernagel J. Kleine Schriften. Bd. 1-2. Gottingen.
Wathelet 1966 - Wathelet P. La coupe syllabique et les liquids voyelles dans la tradition formulaire de l’epopee grecque // Linguistic Research in Belgium / ed. Y. Lebrun, 145-73. Wetteren.
Watkins 1995 - Watkins C. How to Kill a Dragon: Aspects of Indo-European Poetics. New York.
West 1982 - West M. L. Greek Metre. Oxford.
West 1988 - West M. L. The rise of the Greek epic // JHS. Vol. 108. P. 151-72.
West 1997 - West M. L. The East Face of Helicon: West Asiatic Elements in Greek Poetry and Myth. Oxford.
West 2007 - West M. L. Indo-European Poetry and Myth. Oxford.
Wetzels 1985 - Wetzels W. L. The Historical Phonology of Intrusive Stops: A Non-Linear Description // The Canadian Journal of Linguistics. T. 30. P. 285-333.
Willy 2003 - Willy A. xai: mykenisch oder nachmykenisch? // Glotta. Bd. 79. S. 224-48.
Zygis 2010 - Zygis M. Typology of Consonantal Insertions // ZAS Papers in Linguistics. Vol. 52. P. 111-40.
Summary
B. P. Maslov. The metrical evidence for pre-Mycenaean hexameter epic reconsidered.
This article disputes the currently dominant view that the forms avBpoT^ra, avBpsKpovTy, aftfSgOTa^O^ev, a^gorn, attested in the Iliad, are to be scanned with a vocalic /r/. In light of the fact that this sound was lost already in Mycenaean Greek, such a prosodic archaism would point to the preservation of pre-Mycenaean linguistic elements in the text of Homer. Two alternative approaches to the problem are discussed (one positing the omission of the nasal, the other applying N. Berg’s theory of a recent Aeolic substrate in Homer’s dactylic hexameter). Seeing that none of the available explanations of the prosodic anomaly is satisfactory, the article seeks to demonstrate that to account for it we need to posit reduced articulation of the so-called epenthetic consonants /d/ and /b/.
