Why do Epic poets say θάρσος ‘courage’, Athenians θάρρος ‘courage’ and θράσος ‘audacity’, while Alcaeus sings θέρσος? How and when did the four main dialects of alphabetic Greek originate? How and when did the hexametrical Kunstsprache arise? These questions and many more van Beek mulls over in his encyclopedic treatment of syllabic liquids in Greek, a work whose main aim is to establish the regular reflexes primarily of Proto-Indo-European *r̥ and secondarily *l̥ in all dialects of Ancient Greek (Mycenaean included). His coverage far exceeds that stated aim. This wide-ranging and ambitious work, thoroughly revising the author’s Leiden dissertation of 2013, serves as a window onto knotty questions of linguistic chronology. Burgeoning with new ideas, scores of etymological proposals, and discussions of sound laws and literary prehistory, this monograph deserves to be read by anyone interested in the Ancient Greek dialects and the formation of Epic language. Van Beek’s writing, dense in star-spangled forms, is intended for the heavy-duty Indo-Europeanist, as befits a volume published in the series “Leiden studies in Indo-European”; however, it would be a pity if his research proved overly forbidding for Classicists, who would then miss his significant studies of Greek words and texts. Indeed, a joy of reading and a major strength of the book is the author’s “fresh examination of philological data” (498). His study offers plenty of rewarding readings of ancient texts and I will highlight some of his most insightful discussions below; however, I also have serious reservations. I will outline the structure of the book, discussing in detail what seems to me its most problematic idea, namely, a special outcome of “Epic *r̥” (ch.6).
Chapter one surveys his terrain. Chapter two begins his chronological treatment with the Mycenaean reflexes of *r̥. Van Beek focuses on the signs spelled <Co-> (not <Co-ro->), representing either an outcome -or- or, more daringly, a preserved -ṛ-, but “certainly not ro” (100). A proposal for a preserved r̥ is certainly on the books, associated most prominently with Heubeck, but this reading has not been widely followed; van Beek provides fresh arguments in its favor. In the alphabetic dialects (ch.3), he proposes that *r̥ regularly yielded -αρ– in Attic-Ionic (καρτερός, τέταρτος, Att. καρδία, etc.) with the vowel slot preceding the liquid, which is highly significant in the author’s view. This claim also entails that all instances of –ρα– in Attic-Ionic (e.g., ἔδρακον, κράτος, θρασύς etc.) be explained either as analogical or as borrowings from Epic, more on which below. The same outcome holds for West Greek (Doric), though he posits a novel development for Cretan: *r̥ > -αρ- but -ορ- after labials. In all Aeolic dialects, the regular reflex was -ρο- without further conditioning, e.g. στρότος ‘army’ and ἀμβροτην ‘to err’ (inscriptional forms he judiciously leaves unaccented). This reflex would constitute “an important point in favor of reconstructing Proto-Aeolic” (501), a contested entity. Arcadian has ‑ορ- while Cyprian is deemed uncertain (all dialect outcomes are helpfully tabulated on p.502 tab.27).
Chapter four treats reflexes of liquids in the thickets of “Caland” morphology, a major topic of current research in IE studies. Articulating the complex divisions between phonology and morphology, van Beek finds that in general a-vocalism from *r̥ spreads across related derivatives (so-called “Caland systems”). This vocalism is largely based on the zero grade of the positive adjective, e.g. s-stem θέρσος in Alcaeus is a relic, replaced by θάρσος (Hom.+) based on *θαρσύς ‘daring, confident’ from the zero grade*thr̥s-ú-. In this case, the base form is found only in derivatives such as the verb θαρσύνω and the adjective θαρσαλέος, which must have ousted the base u-stem. So too in the compounds where e-grade -θερσης is replaced by -θαρσης (πολυθαρσής etc.), again owed to *θαρσύς. Van Beek is surely right and the principle of analogical influence proves vital to accurately distinguishing inherited vowel grades from analogical replacements. Chapter 5 is wholly devoted to disentangling the morphology of one root: Greek κρατ-/καρτ- (κρατύς, καρτερός, κάρτιστος, etc.).
Chapter 6 begins his discussion of Epic Greek. I delve more deeply into this chapter because it highlights the author’s skills in careful philological sifting but also underscores the shortcomings of his theoretical framework. He begins by signaling the prosodic (metrical) behavior of inherited stop plus liquid clusters, traditionally Muta cum Liquida (hereinafter McL). In earliest Epic such sequences scanned almost always heterosyllabically, rendering the first syllable heavy, e.g. πᾱτ.ρός (Il.1.396; the dot represents the syllable division). This scansion is assumed to be inherited because it is also standard in Vedic metrics. Epic does, however, allow both treatments: the cluster -κλ- makes position in the vocative Πάτ.ρο̄κ.λε (Il.24.592) but doesn’t in Πάτ.ρο̆.κλε (Il.19.287). Van Beek charts the trajectory of McL scansion within Epic, showing a budding license on the rise: incipient in the Iliad, increasing in the Odyssey and in Hesiod’s poems, it reaches its Epic apogee in the Homeric Hymns (254). To all appearances, then, we are witnessing a diachronic shift. Furthermore, van Beek tracks the change from word-initial position into morpheme boundaries, first at compound seams, e.g. ἀλλό-θροος or πρωτό-πλοος (255; he is following important observations by Sommer, and see as well West 2007: 47 n.66).
As is well known, between Epic and Attic the McL clusters shift towards a tautosyllabic scansion, rendering the first syllable light (πᾰ.τρός), known as correptio attica. This syllabification becomes the norm in certain registers of Attic drama “and may be presumed to reflect the spoken colloquial of the time” (Allen 1973: 211, a work not found in the bibliography).[1] Linear B writing may possibly provide a very early indication of the same syllable division, though precisely what these rules and strategies encode remains disputed. On one account, influential with linguists, Linear B spellings of stop plus liquid do reflect syllable divisions: “tautosyllabic assignment of stop-liquid clusters can be taken for granted” (Steriade 1982: 334). Accordingly, as ti-ri-po-de spells tri.po.des, so medial me-re-ti-ri-ja might spell me.le.tri.yai ‘millers’ (fem.pl.); cf. Miller (1994: 16–17), and with more examples Melena (2014: 110–13). I mention this body of evidence because van Beek does not, and yet it arguably bears on the question of McL scansion. Traditionally, McL is understood as arising with words that couldn’t otherwise be adapted to the hexameter (e.g., Ἀ.φροδῑ́τη), a metrical need that arguably squared with the ongoing shift in syllabification. That is, bards needed to slot in names like Ἀφροδίτη and local audiences were increasingly accustomed to hearing both Ἀ.φροδίτη and Ἀφ.ροδίτη. I bring up this business of McL in relation to changing onsets of medial syllables because here van Beek proposes a radically different account and does so without sufficiently considering research into the syllable.[2]
What is van Beek’s radical new proposal? Crucially for van Beek, this “tautosyllabic scansion is structurally applied only in a limited set of lexemes” (242). That set, once pruned by van Beek, reflects only *r̥. Thus, van Beek locates the origins of McL scansion in the reflex of syllabic *r̥; this metrical argument, in turn, leads van Beek to his heftiest claim: “the most attractive explanation would be that *r̥ was retained within Epic Greek for a considerable period of time after its elimination from the vernaculars, perhaps until one or two generations of poets before the composition of the Iliad” (241). After this prolonged retention, he posits a “subsequent vocalization *r̥ > -ρα- (-ρο- after a labial consonant) that was specific to Epic Greek” (241). Or put differently, van Beek claims that Epic Greek “underwent its own conditioned sound change” (260), i.e. a sound change not found in any dialect. That *r̥ might have been retained longer among the Homeridae is not what makes this thesis so problematic; linguistic changes in the dialects surely did outstrip change within the Kunstsprache. But when van Beek proposes a conditioned sound change in the poetic language, one at odds with all known dialects, I hesitate. I know of no other examples, nor does van Beek offer a valid parallel.
Epic Greek, heir to a long oral tradition and constituted of heterogenous dialects, does not exist on a par with the vernacular dialects. The singers of epic traditions did not learn Epic as a first language, as they would have learned, when children, e.g., East Ionic. Phonological change happens when a child, during language acquisition, acquires a phonology distinct from their primary linguistic data (a standard position in historical linguistics; see, e.g., Hale 2007, esp. ch.1 and part II). A poetic language, such as Epic, is not learned as a first language, and therefore will not take part in the same range of language changes. The point is, of course, critical. Van Beek is aware of the problem, admitting that “[t]hus far, however, no instances of artificial phonology have been identified” (49), by which he means that we do not have instances of phonological changes that took place only in Epic. Later on, he circles back to the problem, but does not raise, so does not address, this criticism, stating only that “[o]ne might ask whether it isn’t far-fetched to posit a special epic reflex -ρα-” (263). He poses a tangential question, whether “Epic Greek was subject to changes in pronunciation” as it was “recited and pronounced” (263). The question, however, isn’t, whether Epic texts changed in pronunciation, but did, or even could Epic undergo phonological changes. Mine is no quibbling detail: his edifice is built on this foundation.
He parallels his phonological change only with the pronunciation of Sanskrit r̥ in later Indic, but this parallel doesn’t hold. Sanskrit r̥ came to be pronounced [ri] in many traditions, as in the very name Sanskrit (Skt. saṃskr̥ta-) or Rig-Veda (Skt. r̥g-), though Old Indic r̥ became in the Middle Indic dialects variously a, i, or u, mostly without trace of the rhotic (e.g., Skt. kr̥ta- ‘made’ > Pāli kata-). Thus, as van Beek rightly states (264), r̥ was “retained in traditional recitations of Sanskrit” mantras and prose but altered according to the regional canons. However, it is not the case that a poetic Kunstsprache of Sanskrit, such as Rigvedic, underwent conditioned phonological changes of the kind van Beek proposes for Epic Greek. We have no evidence, then, for a phonological change in a poetic language, neither in Greek, nor in Sanskrit, nor in any other tradition I am aware of. Likewise, we lack any evidence for a dialect of Greek that did undergo van Beek’s sound change and could have then retrojected the –ρα-/-ρο- sounds back into Epic, as the Middle Indic dialects injected the vowels i, u, etc. (van Beek doesn’t raise this possibility, which actually seems more plausible, though still far-fetched). Accordingly, on this point, his edifice wobbles.
Unfortunately, if he can’t maintain his proposal for an Epic *r̥, many claims throughout his work are vitiated. Specifically, every case that he attributes to Epic *r̥, which is every case of Attic-Ionic -ρα- stemming from a root that once had a syllabic liquid and does not owe its outcome to analogy, must be rethought. He is forced to make the eyebrow-raising claim that words normal in the language of prose authors in fact derive from Epic, e.g. θρασύς, στρατός, τράπεζα. In turn, words which undergo McL will be assumed to reflect *r̥, an insistence that tangles van Beek up in knots. For instance, the metrical evidence of McL leads van Beek to the bold claim that the two inflections of Κρονίων (-ιωνος and -ιονος genitives) “are originally two distinct lexemes” (348). If one does not share the assumption that sequences undergoing McL must reflect *r̥, then, as van Beek himself anticipates, “[t]his conclusion may come as a surprise” (348). He concludes that his scenario “may look overly complicated, but I feel that the metrical evidence asks for such a drastic solution” (350). But I doubt that many will share the feeling.
The McL facts alone do not obviously necessitate so costly a hypothesis. Van Beek mentions but dismisses the importance of shifting syllable boundaries: “Homer also uses the McL license in other lexemes [than those that historically reflect *r̥, JL] and this may well be related to a concomitant change of syllabification in spoken Ionic at the time of composition of the Iliad and Odyssey; but this is not of our direct concern here” (259). Van Beek misses that in the history of Latin we have a solid parallel for shifting syllable boundaries in TR clusters (T = stop, R= liquid). Inherited T.R syllabification is retained in prehistoric and presumably in Very Old Latin. For example, the second vowel e in genetrix must come from the closed-syllable development of short *a and *e to e, which implies that it had been syllabified in its earlier form as *ge.nat.rix (contrast genitor from *ge.na.tor in an open syllable). In Plautus, .TR syllabification has become standard (pa.trem etc.); in Classical Latin, both scansions were permitted.[3] This complicated history within Latin illustrates how McL clusters can shift metrically as the clusters shift phonologically. The parallel with Greek seems clear. As Miller (2014: 306-7) put the matter, a shift in syllabification “permitted poets to exploit the two possible scansions for metrical reasons.” Thus, to squash a divine name into the right metrical sedes, poets expanded a variant available in their native phonologies; maybe the syllabification Ἀ.φροδίτη sounded audacious at first, much as pa.trem might have struck Latin audiences. Van Beek is right that reflexes of syllabic *r̥ make up one major ingredient in the development of McL in Greek hexameters, and I agree with him that previous accounts relying on prehistoric sandhi are far from compelling; however, the costs of his alternative hypothesis are too substantial. I have difficulty seeing why we should assume anything more than meter and syllabification for McL in Greek.
Chapter 7 takes on “Epic Forms with -ρο-”, especially those stemming from *r̥, including several words central to the poems, e.g., βροτός (< *mr̥tós, cf. also ἄμβροτος, ἀμβρόσιος). Aeolic and Achaean origins are the usual explanation for the o vowel, but van Beek argues for a “regular reflex of Epic *r̥ after labials.” In certain respects, this is van Beek’s best chapter — original, insightful, thorough — but in other ways it is flawed as it continues the problematic thesis of the preceding chapter. However, even if you disagree with his “Epic *ṛ,” you can still learn much from discussions of individual words. For instance, van Beek launches into a successful analysis of a word much disputed of late: ἀνδροτῆτα. Earlier scholars (e.g., West, Watkins, both following Wackernagel) took this word back to *anr̥tā́tṃ and saw in it a remarkable preservation of a prior syllabic liquid. Not coincidentally, this word is crucial to our Iliad, which centers on the mirrored deaths of Patroclus and Hector, both souls flying to Hades bewailing their fates, having left behind their ἀνδροτῆτα (Il.16.856-7; 22.362-3). In the sole other occurrence of the word, when Achilles mourns, he longs for the ἀνδροτῆτα of his friend Patroclus (Il.24.6). Some recent proposals seek to brush the word aside, but they fail to explain how attested ἀνδροτῆτα got into our texts. Stepping onto this disputed turf, van Beek yet offers something new by revisiting in granular detail the relevant passages. Following a close reading of the ἀνδροτῆτα passages and its related formulaic nexus, van Beek proposes a meaning ‘manly vigor’ for this word. He argues compellingly for the retained syllabic *r̥ in the prehistory of the Iliad tradition, though he improves on previous accounts that insist the whole line must be of a Bronze Age origin. Van Beek derives the preform *anr̥-tāt- from an earlier stage when “*h2ner- or its reflex could still be used as an adjective meaning ‘vigorous’” (326).[4] He considers the regular outcome to have been *ἀνδρατῆτα (cf. ἀνδραφόνος, ascribed to Solon, with Watkins 1995: 330, 498), but the noun has been remade after compositional forms in ἀνδρο-. In terms of metrics, morphology, and semantics, van Beek’s is the best of recent accounts.[5]
In the remaining chapters, van Beek assesses zero-grade reflexes in thematic aorist stems (e.g., δρακεῖν < *dr̥k̂-e/o- etc.), collects “Remaining Issues Concerning *r̥” (ch.9), addresses *l̥ in a brief chapter where he posits that the Attic-Ionic outcome appears to be -λα- (remarkably, as it goes against his proposal of *r̥ > -αρ-). Chapter 11 provides a relative chronology; chapter 12 summarizes and evaluates results. The work is rounded off with a bibliography and helpful indices.
To conclude, van Beek cuts through vast swaths of philological underbrush to arrive at a cleaner presentation of the data. His book, bristling with proposals, will be the point of reference for all future discussions of syllabic liquids in Ancient Greek. But that modest goal doesn’t describe the riches these 500-plus pages contain. The author’s passion for Homeric poetry and for the close study of the Greek dialects shines through, qualities too often lacking in severior linguistic discussions. I disagree with the author on his central proposals, in particular his flawed notion of an “epic r̥”; but all who are interested in questions of historical linguistics pertaining to Ancient Greek, from the genesis of the dialects to the transformations of Epic, should consult this work.[6]
References
Allen, W. Sidney. 1973. Accent and Rhythm. Prosodic Features of Latin and Greek: A Study in Theory and Reconstruction. Cambridge: Cambridge University Press.
Barnes, Timothy G. 2011. “Homeric ΑΝΔΡΟΤΗΤΑ ΚΑΙ ΗΒΗΝ.” The Journal of Hellenic Studies 131: 1–13.
Bozzone, Chiara. 2022. “Homeric Formulas and Their Antiquity: A Constructional Study of ἀνδροτῆτα καὶ ἥβην.” Glotta 98 (1): 33–67.
Byrd, Andrew Miles. 2015. The Indo-European Syllable. Leiden and Boston: Brill.
Cooper, Adam I. 2015. Reconciling Indo-European Syllabification. Leiden and Boston: Brill.
Devine, Andrew M., and Laurence D. Stephens. 1994. The Prosody of Greek Speech. Oxford: Oxford University Press.
Fortson, Benjamin W. Review of Devine and Stephens 1994, in BMCR 1995.10.09.
— — —2008. Language and Rhythm in Plautus: Synchronic and Diachronic Studies. Berlin: de Gruyter.
Hale, Mark. 2007. Historical Linguistics: Theory and Method. Malden, MA / Oxford: Blackwell.
Maslov, Boris. 2011. “The Metrical Evidence for Pre-Mycenaean Hexameter Epic Reconsidered.” Indoevropejskoe Jazykoznanie i Klassičeskaja Filologija XV: 376–89.
Melena, José. 2014. “Mycenaean Writing.” In A Companion to Linear B: Mycenaean Greek Texts and their World, edited by Yves Duhoux and Anna Morpurgo Davies, 3:1–186. Louvain-La-Neuve / Walpole, MA: Peeters.
Miller, D. Gary. 1994. Ancient Scripts and Phonological Knowledge. Amsterdam / Philadelphia: John Benjamins.
———. 2014. Ancient Greek Dialects and Early Authors: Introduction to the Dialect Mixture in Homer, with Notes on Lyric and Herodotus. Berlin: de Gruyter.
Nussbaum, Alan J. 2022. “Derivational Properties of ‘Adjectival Roots’ (Expanded Handout).” In Zurück zur Wurzel: Struktur, Funktion und Semantik der Wurzel im Indogermanischen. Akten der 15. Fachtagung der indogermanischen Gesellschaft vom 13. bis 16. September 2016 in Wien, edited by Melanie Malzahn, Hannes A. Fellner, and Theresa-Susanna Illés, 205–24. Wiesbaden: Reichert.
Pike, Moss. 2011. “Latin -tās and Related Forms.” Ph.D. Thesis, University of California, Los Angeles: UCLA Program in Indo-European Studies.
Steriade, Donca. 1982. “Greek Prosodies and the Nature of Syllabification.” Ph.D. Thesis, Massachusetts Institute of Technology.
Watkins, Calvert. 1995. How to Kill a Dragon: Aspects of Indo-European Poetics. Oxford: Oxford University Press.
Weiss, Michael. 2020. Outline of the Historical and Comparative Grammar of Latin. 2nd edn. Ann Arbor / New York: Beech Stave Press.
West, Martin. 2007. Indo-European Poetry and Myth. Oxford: Oxford University Press.
Notes
[1] Devine and Stephens (1994: 35), but see the cautious appraisal by Fortson 1995, with his footnotes 13 and 14 on this point.
[2] Studies such as Steriade 1982 or Byrd 2015, or indeed more familiar items in Classics like Allen 1973, play no role and are not found in his bibliography, while weighty works like Devine and Stephens 1994 receive at most cursory mentions. Cooper 2015 (esp. ch.4) treats medial consonant syllabification in fine-grained detail; Van Beek cites this monograph only in an early footnote (15 n.44) as “Cooper 2014” (with the huge range of “chapters 6-10”, i.e. about 150 pages!), which work is then dropped from the bibliography.
[3] Fortson (2008: 20-1, and esp. ch.7) offers a nuanced discussion of Plautine scansion. On this admittedly complex back-and-forth syllabification, Weiss (2020: 77–8) provides a particularly informative section.
[4] An oft-repeated assertion holds that nominal abstracts in *-tāt– (Gk. -τητ-) are derived from adjectives and therefore ἀνδρ(ο)-, occurring solely as a noun ἀνήρ, cannot be the basis of ἀνδροτῆτα. However, this reasoning overlooks the evidence for an earlier adjectival use of PIE *h2ner-, as discussed in detail by Pike (2011: 166-76, 220-1), whom van Beek follows. Of interest among post-van Beek publications, Nussbaum (2022), an earlier version of which Pike cites, demonstrates the adjective-hood of second compound-members in -ηνωρ, which derives from the same root that produces ἀνδρο-. He argues that such compound-members may have the abstract meaning ‘strength, manliness, courage, mettle, high spirits’, citing the “unforced reading” of ἀγήνωρ ‘having great manliness, courage’, ἀνήνωρ ‘cowardly’, and ὑπερήνωρ ‘overweening’. Nussbaum’s research is compelling and thus corroborates proposals, such as that by Pike and van Beek, which derive ἀνδροτῆτα from earlier *anr̥-.
[5] Bozzone (2022) was published too late for van Beek to refute, but her account is ably anticipated. Bozzone plumps for the weakly attested variant ἁδρότητα [sic] ‘vigor, ripeness,’ finding in its favor that it appears in later authors, is morphologically transparent, is metrically preferable, and offers a better semantic reading — “[i]n short, it could easily have been generated by the synchronic grammar of the poet.” This is, however, not so much a point in its favor as the definition of a lectio facilior. And as van Beek rightly states, “[a]n insurmountable objection, finally, is that ἁδροτῆτα is the lectio facilior” (322 n.94). Van Beek does not mention the untenable account by Maslov (2011), who proposes a permitted scansion ἀ.ν(δ)ρο.τῆ.τα. He discusses and refutes Barnes (2011).
[6] I am grateful to colleagues who helped me think through the topics in this review: Stephanie Jamison, Elisa Migliaretti, Brent Vine, and Anthony Yates, none of whom are responsible for the views I’ve expressed.