Steve Reece, Homer's Winged Words: The Evolution of Early Greek Epic Diction in the Light of Oral Theory. Mnemosyne Supplements 313. Leiden/Boston: Brill, 2009. Pp. xi, 413. ISBN 9789004174412. $241.00.
Reviewed by D. M. Goldstein, University of California, Berkeley
Time was when we called an apron a napron, a newt an ewt, and a nickname an ekename. These changes all illustrate the phenomenon of resegmentation,1 which has shaped the lexicon of many languages of the world to varying degrees. In his new book, Steve Reece explores the impact such changes have had on the Homeric Kunstsprache. Taking a cue from Leumann's Homerische Wörter, Reece argues that resegmentation played a key role in the creation of the Homeric lexicon (summed up in the slogan "Homeric diction begets Homeric diction"), and explores how the phenomenon can illuminate the hitherto opaque histories of certain words. Where Reece diverges from Leumann is in the mechanism driving such change: for Reece, resegmentation is a decided byproduct of the oral context and transmission of the epic (whereas Leumann did not always make explicit claims about the mechanism(s) driving resegmentation; see p. 9 n. 9). In particular, he is interested in cases of resegmentation that took place at what he calls the bardic stage (early second millennium BCE, and perhaps earlier, to the middle or late eighth century BCE) and the transcriptive stage (middle to late eighth century BCE) of the epic.
This is an exciting topic, and one about which we know very little, since it has received only sporadic attention in the past. There is indeed no comparable work. Reece's book will be of interest to those interested in linguistic change and the history of Greek, prosody (in the broader sense encompassing e.g. rhythm, intonation, and syllabification, and not just metrics), and of course the Homeric Kunstsprache and the oral-formulaic theory of performance and composition. We are indebted to Reece for undertaking such a comprehensive investigation of the topic (the book is just over four-hundred pages).
The strengths of this book are several. In addition to marshalling a wide spectrum of examples of resegmentation, Reece also adduces a wealth of evidence from ancient sources such as scholiasts and lexica. While historical linguistics and reconstructed forms play a role at various points in the book, technical language is kept to a minimum, so the book should be accessible to a wide audience; its readership is, however, narrowed somewhat by the practice of translating Greek quotations only rarely. The primary weakness of the volume is that a number of its claims are too speculative, and the arguments are not as precise as one would like. Of course given the nature of the phenomenon and often paltry data, some speculation is unavoidable, but there are measures Reece could have taken to put his analyses on firmer ground. In particular, it would have been helpful if he had spelled out in greater detail the benefits of his model over competing explanations. For, as it stands, it is not always easy to see why we should place stock in resegmentation in the face of alternative explanations. Further issues are raised below.
The book consists of nineteen chapters divided into two parts. In part one, Reece begins with a general introduction (chapter one), then presents a brief cross-linguistic survey of resegmentation (chapter two), with a particular focus on Middle English. This is a very informative and extremely interesting collection of data. From here, he turns to the situation in Homeric Greek (chapter three), and presents cases of resegmentation that arise from nu-ephelkystikon and final-nu (chapter four); moveable and final sigma (chapter five); moveable and final kappa (chapter six); vowel elision (chapter seven); and finally a chapter on the reanalysis of Homeric toponyms (chapter eight).
Part two contains case studies of individual cases of resegmentation, in particular the ethnonyms Ἄβιοι and Γάβιοι, and Σελλοί and Ἑλλοί; the particles ταρ and ἄρ'; as well as handful of other cases, such as: φολκός and (ἐφ)ὁλκός; γνυ- and ἰγνυ-; κατὰ κρῆθεν and κατ' ἄκρης; ἀσφοδελός and σφοδελός; ἀσάμινθος and νασάμινθος; σῶκος and ὠκύς; πτολέμοιο γέφυραι and πτολέμοιό γ' ἔφυραι; πτερόεις and ἄπτερος. The book concludes with a section of illustrations and maps; a large dose of appendices (pp. 345-361); bibliography; index locorum; index nominum; and index verborum. There is no index rerum.
I call attention here to four general problems that affect the book as a whole; in the next section, I comment on specific portions of the text. The first, and perhaps most significant, deficit is that Reece does not summarize the conditions under which resegmentation is more or less likely to occur. To be sure, he does offer in chapter two a very informative cross-linguistic overview of the phenomenon, with various comments in this direction. At the end of this, however, what I was hoping to find was a summary of features that characterize resegmentation, which could then be used as a guide to analyzing potential cases of the phenomenon in Greek. Other treatments of resegmentation indeed attempt to do this,2 but given the scale and detail of Reece's study, he is in a unique position to offer new insights into this phenomenon.
In particular, the discussion could have been improved with greater attention to the following conditioning factors:
1. Prosodic domain: there is a definite tendency among the data in Reece's cross-linguistic survey for resegmentation to affect proclitics, for instance.
2. The syllable: are there any properties of the syllable that make resegmentation more or less likely?
3. Frequency: one might reasonably expect a resegmented word to occur more frequently in a particular collocational frame than on its own. Indeed, given the role that frequency can play in language change generally, this is an issue of considerable importance, and more should have been said on it throughout the book.3
It would also have been helpful to discuss the differences between the onset-to-coda (a napron > an apron) and coda-to-onset (an ekename > a nickname) change (as well as cases of degemination at word-boundary), as well as the effects (if any) of segmental transition probabilities, syntactic structure, and speech style on resegmentation.
To be sure, Reece is not unaware of the above issues, but they are not systematically addressed. For instance, when possible, Reece does indeed note likely formulaic frames that may have contributed to resegmentation. At other times, however, a word is of such scant attestation that the evidence does not suggest a particular source phrase for resegmentation. It is precisely in these cases that it would have been useful to have a dossier of tendencies to rely on when trying to decide whether resegmentation occurred or not.
The second issue results from the first, and pertains to the nature of Reece's explanations as to why a particular case of resegmentation took place. It seems reasonably clear (although it would have been helpful to have this stated more explicitly) that Reece thinks of resegmentation as often rooted in misperception, and that this misperception results from the "oral/aural interplay of bardic performance" (p. 331, passim). Indeed, misperception seems prima facie to be a very likely explanation for a great many cases of resegmentation (although whether or not it is misperception specifically in a poetic context is a separate issue, which I touch on below). But one has to be careful with this explanation. For one imagines that, in the case of rare words, a particular hearer (whether bard or audience member) might not have well-entrenched phonological representations. In this case, misperception may well be enough for resegmentation. If, however, a hearer does have a strong representation of the word, then misperception would seem to be a necessary but not sufficient cause for resegmentation. For cases of this latter type, any explanation will then need two components: misperception, combined with some other mechanism. At times, Reece does invoke misperception in combination with other operations, such as analogy (p. 229, 333). But with the addition of the second mechanism, one wonders what role misperception should play at all. So for instance on p. 278, Reece describes the reanalysis of νασάμινθον as ασάμινθον as "almost inevitable" given its presence in (presumed) phrases such *ἀργυρέαν νασάμινθον. But then he adds that an aversion to superheavy syllables would have provided further motivation. Fine, but why is this not just a case of avoidance of superheavy syllables then? The issue of multiple-causation is tricky enough that it should have warranted its own discussion.
Furthermore, and this was noted above, Reece rarely suggests or defends his claims against competing explanations. In my view, this addition would have strengthened the book considerably, because it would have made clear why resegmentation is a better explanation for an opaque form than others. But as it stands, in several of Reece's discussions, I found myself wondering whether other mechanisms might not be at work, such as sporadic sound change; dialect borrowing; phonaesthesia/sound symbolism;4 or folk etymology (consider for instance the addition of the initial h in the English word hangnail5).
Third, there is some confusion as to whether resegmentation is purely a literary phenomenon (taking place within the confines of the epic performance) or one taking place more broadly across speech communities at large. In the first part of the book, Reece seems to allow for both; note e.g. p. 8: "I propose that during the period of oral transmission acoustic uncertainties, especially regarding word boundaries, were continually occurring, just as they occur to this day in everyday speech: a bard uttered one collocation of words, but his audience--which often no doubt included an aspiring bard--thought it heard another." I take this to mean that the bard initiates the change, but the spread takes place within the speech community. Elsewhere Reece identifies an error as occurring specifically in bard-to-bard transmission (p. 213): "Probably long before the time of Homer one of his bardic predecessors, who had never actually been to distant Dodona, on the fringes of the Hellenic world, had misheard a formulaic expression in which the name was embedded and consequently metanalyzed Ἑλλοιas Σελλοί." Resegmentation can of course either be an inner-epic phenomenon or one that takes place in the vernacular, and disentangling these alternatives is a matter of considerable difficulty, of which Reece is not unaware (see his remarks on p. 155). But without some way to diagnose metrically induced versus non-metrically induced cases of resegmentation,6 it does not seem prudent to assign a change to either category. Of course one could raise the point that, since Greek is by and large attested only in metrical texts at this stage, we are not in a position to know what is happening in the domain of daily communication. Fine, but this point cuts both ways: if you do not know what changes are taking place in the vernacular, then how can you claim that the resegmentation witnessed in Homer results solely from the epic tradition and setting?
Finally, even if we were to accept all of Reece's analyses, he overstates the role of resegmentation in the creation of new lexical items. For instance, he writes on p. 70: "metanalysis at word boundaries was a productive force in the creation of new forms and words in the pre-Homeric Kunstsprache." (Compare similar remarks on pp. 26, 60, 3297.) While I think Reece is right to advertise the phenomenon so that it receives the attention it deserves, it nevertheless seems odd to characterize resegmentation as a "productive" word-formation process. It is true that productivity is a question of degree, but still I at least do not see enough regularity (or frequency) in the phenomenon to warrant this description. I also wondered about the implications of accepting Reece's overall model. For instance, what about resegmentation due to scribal misperception: if we accept Reece's account, how does this phenomenon fit into the picture?
pp. 15-26. In the cross-linguistic survey of resegmentation, reference could profitably be made to the Lycian nasal preterite,8 as well as the Vedic Sanskrit form naveda- 'observing, cognizant' from veda-.9
p. 9. On collocational chunks in language production, see the insightful article by D. Bolinger, "Meaning and Memory," Forum Linguisticum 1 (1976): 1-14.
p. 13. For discussion of the notion of a "word," see A. Morpurgo Davies, "Folk-linguistics and the Greek word," in G. Cardona and N.H. Zide, edd., Festschrift for Henry Hoenigswald on the occasion of his seventieth birthday (Tübingen 1987): 263-280.
pp. 199, 227, 235. Reece has a habit of describing particles like γε as "ubiquitous" (the adjective also shows up on p. 315 n. 1 in reference to the formula ἔπεα πτερόεντα). This characterization has a deleterious effect on his analyses, because it essentially allows him to treat particles as ad hoc phonetic filler without motivating their appearance. It may indeed be the case that some particles served as phonetic filler in a line (so Sanskrit eva, according to some), but such a claim needs to be defended, and not casually assumed.
pp. 351-358. Reece offers an interesting appendix on the phenomenon of mondegreen, which is typically defined as a misparsing of a phrase or line from a song lyric. (A well-known example is from Jimi Hendrix' song "Purple Haze," where "excuse me while I kiss the sky" is parsed as "excuse me while I kiss this guy.") One feature that Reece should have made clearer is that this phenomenon often produces semantically odd output, but tends not to create new lexical items. Rather, it seems that speakers are doing the best they can to fit a lyric into forms that already exists in their lexicons (and often at the expense of meaning). Indeed, Reece himself acknowledges this point on p. 357, but on the very same page it also becomes clear that he considers Homeric resegmentation a type of mondegreen. In addition to this inconsistency, such a view also assumes that Homeric song is capable of giving rise to mondegreens of the sort that we find in, say, rock music. If so, Reece needed to provide a sketch of the musical/prosodic setting of the epic in support of this idea. On a different note, I occasionally had difficulty understanding why certain items were included in this section, such as blog from web log; for the truncation here is not, I presume, the result of misperception or misparsing.
Typographical errors are few, but not altogether absent. The utility of the book would have been enhanced by an index rerum.
1. The phenomenon is actually known by several different names, including metanalysis, reanalysis, and junctural metanalysis; see the exhaustive list on pp. 359-360. Reece prefers this last term. Despite his defense of the phrase as "the most descriptive and precise" (p. 17), I use the term resegmentation in this review.
2. See e.g. A. Devine and L. Stephens, The Prosody of Greek Speech (Oxford 1994): 234-271. It is disconcerting that this seminal work has such a slight presence in Reece's account.
3. For a recent overview of the role of frequency in language change, see J. Bybee and P. Hopper, edd., Frequency and the Emergence of Linguistic Structure (Amsterdam 2002).
4. On which, see S. Ullman, Semantics (Oxford 1962): 82-92.
5. See E.C. Traugott and R.B. Dasher, Regularity in Semantic Change (Cambridge 2001): 14.
6. For examples of this type from Sanskrit, see T. Oberlies, A Grammar of Epic Sanskrit, (Berlin/New York 2003): xxxv.
7. Reece's account could perhaps be strengthened with a tally of the most likely Homeric cases of resegmentation, which could then be compared to similar counts from e.g. Spanish, French, and earlier stages of English.
8. See A. Garrett, "The Lycian nasalized preterite," Münchner Studien zur Sprachwissenschaft 52 (1991): 15-26.
9. See J. Schindler, "Ein rigvedisches Wort: návedas-," in H. Ölberg and G. Schmidt, edd., Sprachwissenschaftliche Forschungen: Festschrift für Johann Knobloch (Innsbruck 1985): 351-360; B. Fortson, Indo-European Language and Culture (Oxford/Boston 2010): 8.