Two
Comparability of the Documents

If our comparative analyses of the Homeric Odyssey , the Old English Beowulf , and the Serbo-Croatian Return Song as exemplified in the recorded repertoires of the Stolac guslari are to have real validity, we should start by making an attempt to understand the "gross anatomy" of the different texts, that is, the nature of the actual extant documents. In the past it has been customary to assume equivalent authority among the Homeric texts we find whole only in the medieval period, the Anglo-Saxon insular minuscule manuscript of Beowulf from the tenth century, and the Yugoslav epic songs taken down from actual oral performance, whether acoustically or pen in hand by an amanuensis. But quite clearly each of these is a unique medium with a unique history, and before simply erasing their inherent differences and proceeding on to comparative evaluation, it is well to discover what we can about their individuality. The shreds of information about the Homeric and Old English documents are admittedly and frustratingly few, and it will occasionally be necessary to fall back on a best hypothesis where no fact is available; we can, however, learn an appreciable amount merely by struggling toward the sort of full and unambiguous answers that prove so rare in the ancient and medieval periods. And at the same time, not incidentally, we shall be taking another step toward the kind of complex and truly comparative poetics that oral literature studies deserve.

The Journey of the Homeric Texts

The Iliad and the Odyssey , as we have them in the Byzantine recensions that are the earliest complete texts of the Homeric epics, have stubbornly resisted the text and transmission models imposed upon them by generations of scholars. Each set of hypotheses, however apposite in some ways, seems to founder on

at least one important stage of the manuscript record: either the commonly accepted idea of the "vulgate," the phenomenon of the so-called wild papyri of the third century B.C. , or the elaborate editing undertaken by the Alexandrian scholars escapes the web of overall explanation. This is perhaps no wonder, for we are dealing with a history of which we know tantalizingly little and one that stretches from the eighth century B.C. to the tenth century A.D. —that is, nearly two millennia—without a single surviving complete text to anchor supposition. Notwithstanding these severe problems, by bringing to what we can learn about the different stages of the transition a model for textual evolution that suits the likely facts of emergence out of an oral tradition, we can at least approach a coherent view of what our texts of the Iliad and the Odyssey are and how they must relate to Homer's traditional poems.

Although Homer himself is customarily dated to the eighth century B.C. , we have no evidence for a fixed text of his poems until the version assumed to have been compiled during the Panathenaic Festival of the sixth century B.C. While we may choose to agree with Albert Lord's (1953) persuasive argument that the fullness of the poems as we now have them makes it likely that they were oral-dictated texts, that is, that they were taken down from dictation in a manner which allowed the singer to proceed relatively slowly and to compose carefully, we may also imagine the earliest stage of transmission of fixed or semi-fixed texts as primarily memorial.^[1] As Parry contended from his first writings onward, the tradition epitomized in Homer must have been vigorous and continuous for centuries before the monumental poet composed the epics of which some version has survived, and the tradition may be traced back as far as the Mycenaean era.^[2] Likewise, there is no reason to believe that, even if the Iliad and Odyssey that have reached us were fixed in the sixth century B.C. , oral composition of the Homeric ilk immediately ceased.^[3] More likely, the Panathenaic Festival simply provided the opportunity and perhaps the impetus to commit to written and permanent form what Gregory Nagy (1979, 7) has identified as a Panhellenic "standard" version of Homeric poetry, one which "synthesizes the diverse traditions of each local city-state into a unified Panhellenic model that suits most city-states but corresponds exactly to none." If this is so, then we may assume that the more parochial versions of Homeric poetry—now at least two centuries removed from the probable date of composition by Homer—must have continued to disseminate themselves, at first through oral composition and later through memorization and rhapsodic performance.

For available evidence suggests that both oral composition, the craft of the

[1] For discussion of memorial preservation in Middle English literature, see Parks 1986; of related interest is Opland 1975. For an opinion in opposition to Lord's hypothesis, see Hillers and McCall 1976 on Near Eastern traditions.

[2] Cf. Page 1959, 218-96; and Janko 1982. On Indo-European roots, see G. Nags 1974, 1979.

[3] On the lack of effect of fixed texts on a continuous tradition, see Lord 1967b, 1986c.

aoidos , and a creative brand of memorization, the province of the rhapsode (rhapsôidos ), contributed to the early transmission of Homer. In ways we do not entirely understand, the sixth-century Homeridae, or "sons of Homer,"^[4] seem to have followed the great poet in safekeeping his poems; they claimed to recite Homer ek diadochês ("by right of succession"), or almost by a kind of exclusive copyright. Most scholars interpret their relationship to the great poet not as real kinship but as a guild, a cooperative effort to forestall the proliferation of different versions of the poem that must have resulted from the rhapsodes' activities. For alongside the singer of tales, whose primary oral tradition must have begun to recede after Homer, stood the rhapsode, who, in taking as his principal charge the effective oral recitation of Homer from memory or script (often in competition with other rhapsodes), must have felt relatively free to modify the canonical, received work. Indeed, he may well not have realized, or for that matter been prepared to recognize, that such variation was taking place; in the last stages of primary oral tradition, there would be an opportunity for rhapsodes to draw on their own experience of the singer's art to modify phraseology or perhaps even narrative structure. It is telling that references to the Homeridae stem from about the time when, on the one hand, rhapsodic activity and local traditions had produced many variants of what the ancient Greek world knew as "Homer" and, on the other, the Panathenaic text was reputedly being compiled.

This hypothetical first fixed text was apparently a response to What is known as the Panathenaic Rule, a law that governed fidelity to text in rhapsodic competitions from the sixth century B.C. onward, although there also exists a well-known legend that Solon or Hipparchus the son of Peisistratus first "arranged to have all the Homeric poems recited at the Panathenaea in relays of rhapsodes" (Lesky 1966, 73). Whatever the case may actually have been, all of these clues point to the establishment of a fixed text in that time and place. Scholars have explained this text either as a "Peisistratean recension," a hypothesis assembled from reports of interpolations made in the name of local patriotism and from the classical legend (first attested in Cicero) that Peisistratus was himself responsible for the arrangement of the poems,^[5] or as a text brought to Athens by Hipparchus, perhaps directly from the Homeridae of Chios.

As mentioned above, however, Nagy's idea of the gradual evolution of fixed Homeric texts as a function of both composition and proliferation best fits both the surviving evidence and what we know of oral tradition. As he puts it,

composition and proliferation need not necessarily be related as an event followed by a process : the evolution of the fixed texts that we know as the Iliad and Odyssey

[4] For discussion of available evidence on the Homeridae, see Davison 1963a, 235; Davison 1963b, 219.

[5] For a review of ancient sources and opinions on this point, see T. Allen [1924] 1969, 226-38; he discounts the theory of a Peisistratean recension, seeing it as a legend without basis in historical fact.

may be envisaged as a cumulative process, entailing countless instances of composition/performance in a tradition that is becoming streamlined into an increasingly rigid form as a result of ever-increasing proliferation. (1979, 8; see also G. Nags 1986)

Besides taking into account the political aspects of contemporary Panhellenism, this explanation has the advantage of not overlooking the obvious fact that by the sixth century Homeric poetry had already spread from its probable source in Ionia to all parts of the Greek world and, equally important, would continue to be (re-)composed and performed throughout those areas for many years to come. With the hypothesis of a text at this point in history, then, we posit a single epitome: a pair of monumental poems that bear the burden of a long tradition, that represent the Panhellenic fervor of the emerging city-states, and that are spoken (and written down) in a polydialectal Greek no one had ever used for non-poetic purposes. We most assuredly do not posit the immediate demise of the large tradition and the smaller local traditions out of which the poems were fashioned.^[6] The Homeric texts are monumental and no doubt the flower of their now-perished tradition, but they did not exist outside of that tradition.

The Athenian texts of the Panathenaea or some version of them then passed into the capable hands of the Alexandrian scholars, most notably Zenodotus of Ephesus (head of the library in the mid-third century B.C. ), Aristophanes of Byzantium (head 195-180 B.C. ), and Aristarchus of Samothrace (head ca. 180-145 B.C. ). There is considerable historical irony in the fact that we know a great deal about the editing procedures employed by each of these men, even the details of their copyediting marks, but almost nothing about the actual texts they produced, if in fact they did produce whole editions of Homer. Further, since their learned observations and opinions seem to have had little effect on the transmitted vulgate text—a surprising turn of events that may have more to do with the lack of anything approaching a contemporary book trade than with lack of attention to their remarks^[7] —we are left to infer their role in transmission indirectly. Davison (1963b, 223) and others

[6] On this point it is instructive to compare the well-recorded Serbo-Croatian tradition, which consists of the set of all local traditions (among which there is considerable variation), each of which in turn consists of singers' individual or idiolectal traditions. See further the latter part of this chapter, along with chapters 5 and 8.

[7] On the basis of Aristarchus's 664 known readings and their influence, T. Allen ([1924] 1969, 304) concludes that "it is plain we possess no 'Aristarchean edition,' and as the papyri (which are all non-Aristarchean) form a catena from 300 B.C. to A.C. 600, it is also plain that no Aristarchean edition was ever put on the market." Allen is not alone in explaining the situation with the general observation that "at all periods of the Greek world the connexion between organized learning and book production seems slight, if not non-existent" (p. 320); cf., e.g., S. West (1967, 16): "the view that Aristarchus published the Vulgate involves an anachronistic conception of the relationship between the scholar and the book trade: the Museum was not a publishing house; there was no Alexandrian University Press."

note that with Aristarchus's rather conservative but well-founded style of editing Homer, the number of so-called wild papyri, those fragments of the Iliad and Odyssey that vary considerably from the received vulgate, diminishes sharply.

Although silent witness is usually the rule in assessing the contribution of the Alexandrians, we can appreciate the enormity of their task by taking account of the fact that the Library collection of Homeric poetry contained an imposing number and variety of texts. Of the editions current around 300 B.C. , just before Zenodotus assumed command, there are records of thirteen that contain readings adopted by men of letters (kat' andra ), over sixty-six anonymous "city" editions (kata poleis or hai politikai ), and fifty-two vulgate (koinê ) texts. Added to these variants from all over the Greek world, which most scholars assume were collated by Aristarchus but not in any sense "published," is a group of contemporary papyrus fragments heterogeneous enough to be conventionally designated as the "eccentric" or "wild" papyri. We do not know exactly how the Alexandrians treated these excerpts from the Iliad and Odyssey , but we can observe, as noted above, that their number seems to fall off rapidly after Aristarchus. Perhaps S. West's (1967, 17) suggestion that "the post-aristarchean tradition underwent the influence of the Aristarchean Text, without being descended from it," best explains the discernible facts without depending on an anachronistic model of transmission.

But what of this mélange of texts, and especially of papyri? How do they relate to the hypothetical Panathenaic text, itself at least two centuries removed from the supposed time of Homer? Even if we assume with most scholars that the sixth-century Athenian Homer became the vulgate, which in turn gave way to the Byzantine codices on which we base our modern editions, we must still deal with those witnesses that are probably not a part of the direct line of descent. To begin with, we should recall the point made above that a Panathenaic epitome, a fixed text made under the rule for rhapsodic competition, does not in any way presuppose the end of the oral tradition. Rather, it is only logical to believe that the art of oral verse-making continued on for some time, albeit from that point on exclusively in the diverse local traditions rather than Panhellenically.^[8] The fieldwork and collecting of Parry and Lord have shown that the local tradition is a very important unit in the overall ecology of oral epic composition, and on the basis of evidence gathered firsthand in this living laboratory it is clear that just such a great intersectional and rationalizing force as Panhellenism would be required to mold the disparate local traditions into a pair of national poems. More importantly, the making of such a text— and its use as a standard for rhapsodic competitions—would not in any way

[8] It is worth noting that, from the point of view of the history of an oral narrative tradition, such an epitomized text is endemically "unnatural"—that is, it is not the characteristic medium of the tradition but rather a neologism created under political influence.

cause the local traditions from which it was drawn to cease their own independent activity, for by its very nature oral tradition does not recognize the individual text or performance as definitive. At the outset, then, we should realize that it is only natural that the Alexandrians (and the Greek world as a whole) were awash in versions of Homer; the continuing oral tradition and later the "memorization" and "performance" of generations of rhapsodes must have further parochialized the diversity that the Panhellenic spirit sought to counter. Even limited (and chiefly memorial) "publication" of the Athenian epitome through the rhapsodes would lead to variation, since we must expect that such performers often drew from their own Homeric wordhoards to embellish and personalize their performances.

The Ptolemaic papyri present a problem of a different but ultimately related sort. They diverge from the canonical Homeric texts in ways that cannot be explained as purely mechanical corruption associated with scribal practice, and S. West (1967, 11) makes the telling point that although we conventionally refer to them as "wild," "yet there is no evidence that at the date they were written there was anything abnormal about these texts." Indeed, the early papyri arc characterized by a high proportion not of substituted but of additional lines,^[9] and these additions arc not random but of a particular kind:

The proportion depends partly on the context: passages containing many versus iterati ... or a summary of a typical scene described elsewhere in greater detail ... attracted plus-verses, while a passage for which there arc no close parallels elsewhere in Homer was likely to remain free from them. Concordance interpolation exercised a powerful attraction: thus a line or group of lines which follow a particular formula in one place are inserted after it in another passage where they may be rather less suitable. (pp. 12-13)

This "concordance interpolation," as West calls it, belies the existence of a still-fluid vestige of oral tradition, perhaps by this point exclusively the possession of rhapsodes (or even schoolmasters)^[10] who could read and write but had committed much of Homer to memory. However far away we may be from the composing aoidos , we have not yet very nearly approached the stage of a single absolutely fixed, inviolate text. The papyri argue that, even while Aristarchus and his fellow scholars labored over the editing of their manuscripts, others outside the mainstream of textual transmission—but still very much a part of the traditional context—were performing and recording versions of Homer faithful to the overarching tradition yet couched, as our three Yugoslav guslari might put it (see below), in slightly different "words."

In fact this particular kind of interpolation, one that adds lines traditionally associated with other lines or with specific themes or typical scenes, can hardly

[9] West feels that Homeric style was by its very nature always vulnerable to such additions, but that there are even more such lines than even that endemic vulnerability would seem to permit.

[10] See T. Allen [1924] 1969, 271-306.

have been due to any other cause. To put it proscriptively, only those acquainted with the multiformity of Homeric epic tradition could make "errors" of this sort, just as only singers actively involved in the process of composition—as distinct from this latter-day practice of reinterpreting multiforms, mistakenly thought of only as "interpolation" into an early editio princeps —would be able to engender true "narrative inconsistencies."^[11] And once one brings to bear the well-documented example of multiformity in the Serbo-Croatian oral epic tradition, the case is complete: what Lord (1960, chaps. 2-3) illustrates as customary variation among instances of formulas and themes in the epske pjesme of such singers as Ibro Basic^[*] , Halil Bajgoric^[*] , and Mujo Kukuruzovic^[*] is very similar to the sort of "concordance interpolation" West identifies as characteristic of the relationship between the papyri and the canonical text of Homer.^[12] In a real sense the tradition of Homer was, if not active and "primary-oral," at least fluid and responsive possibly as late as the second century B.C.^[13]

From this point on the wild papyri taper off rather suddenly, evidently in part because of Alexandrian scholarship, and perhaps eventually in part because the much more capacious codex, which could entirely contain a long work and thus theoretically help to standardize it, was to replace the papyrus roll, in which little more than a single book of the Iliad or Odyssey could be recorded. The vulgate text now effectively goes underground until its emergence in the first complete surviving Iliad , Venetus Marcianus 454, an early-tenth-century minuscule manuscript that remains the most important source for modern editions because of the famous A Scholia^[14] included with the text. Almost in concert with its mysterious past, the A manuscript then underwent textual criticism by generations of Byzantine scholars and

[11] On narrative inconsistencies, see esp. Lord 1938; Gunn 1970; Foley 1976a. See further chapter 10 below.

[12] It is only fair to note that West (1967, 13), being unfamiliar with the Serbo-Croatian tradition on a philological level, concludes otherwise, namely that "the relatively minor scale of the interpolations argues against the view that there is a connection between the eccentricities of the early texts and the long oral tradition of the poems, except in so far as the rather discursive style suitable for oral technique attracted interpolation." As I have tried to contend, the "relatively minor" differences foregrounded in the papyri are major indicators of oral traditional variation, such variation being marked by its rather conservative limits. Wholesale invention, or even variations that occurred outside of the associations of line or scene that West herself describes, would be counter-evidence against the influence of oral tradition. As Lord puts it (1960, 120), "multiformity is essentially conservative in traditional lore, all outward appearances to the contrary."

[13] For descriptions of a parallel kind of oral transmission, Middle English romance, see Baugh 1959, 1967; H. Duggan 1976.

[14] These consist of versions of Didymus's On the Edition of Aristarhus , Aristonicus's On the Sings of the Iliad and Odyssey , Nicanor's On Punctuation , and Herodian's Prosody of the Iliad ; the A manuscripts thus contain more information than any other single source about Aristarchus's principles of edition.

re-emerged with a number of other texts of Homer in the West. As Davison (1963b, 226) notes, it was rediscovered in the Marcian Library in Venice, "to which it came in the fifteenth century from the library of Cardinal Bessarion, only to be forgotten until the late eighteenth century." Although critics have labored. long and diligently to establish stemmata for the surviving manuscripts from the Byzantine period, most agree that a definitive genealogy for the extant witnesses is beyond our reach. But even if such a genealogy could be established, we would still be faced with about a millennium during which the complete text of Homer is unattested in any direct way. At any rate, the poems reach their true editio princeps in 1488 with the printing of an edition of Homer by Demetrius Damilas. From that point on, and especially after Wolf's Prolegomena of 1795,^[15] the problem of setting the text and deriving its antecedents has turned on the answers made by various ages and schools of criticism to the Homeric Question itself.

In summary, then, as we enter on textual studies of Homer to be undertaken in a comparative context, it is well to recall the long and circuitous journey of the Homeric poems from the eighth century B.C. forward almost three millennia to our own time. The circumstances of the recording of the great aoidos himself remain completely unreported, of course, but we may imagine a written or, more probably, memorial tradition of preservation through the Homeridae^[16] down the two centuries to the Panathenaea and what was probably the first full text. Whether this memorial tradition was based on the kind of oral-dictated text described by Lord we shall perhaps never know, but in any case some variation from the performance of Homer would have occurred no matter how pure the intent of the preservers. The singers and rhapsodes who transmitted the poems must have made their own contributions, according to their lights and their usual mode of performance; as noted above, even the inertia provided by a manuscript text did not prevent a rhapsode from "stitching" together episodes of the poems as he performed them. What surfaced at the Panathenaea, whether under the aegis of Peisistratus or not, must have been a text true in its essential nature not only to the fact of the eighth-century performance but also in some measure to the multiform idiom that allowed the poems to live and flourish after their monumental poet was no more.

Gazing along the probable path of the poems as they reached the first

[15] See further A. Parry 1971, xiv-xvi, xxiv; Foley 1988, chap. 1.

[16] One possibility is that urged by M.W.M. Pope (1963, 6-7): "What we must have then to account for the variant reading is different dictations from an originally single version . This version one must imagine not as a written text but as a memorised one, learned from the lips of the poet (a possibility envisaged by Parry), and transmitted at first by singers, who would naturally retain for some time their capacity for original composition, and later perhaps by pure reciters." While I prefer Nagy's model of a Panhellenic fixation of the text as sketched above, Pope's comments offer a realistic explanation of the post-sixth-century situation.

professional textual critics in Alexandria, moreover, we realize that nowhere in the ancient records is there incontrovertible evidence that the vulgate and Panathenaic texts were one and the same. It is of course convenient to assume a direct line of descent, since we can then avoid having to deal with problems of "genuine" or "spurious" texts. But by making that assumption we obscure if not forget a crucially important dimension of Homeric art; we push aside as an unnecessary complication the evidence that no less than 131 separate editions of Homer were deposited in the Alexandrian Library, some with commentary, some as part of the vulgate record, and some attributed to various cities. When we add to this evidence the phenomenon of "concordance interpolation" that characterizes the surviving Ptolemaic papyri, the picture comes clearer. The practice of Homeric poetry must have been alive in some form all over the Greek world for centuries after Homer;^[17] by insisting on a simple, literary model for transmission of the Homerid, Athenian text directly to Alexandria and beyond we miss that essential point. The Iliad and the Odyssey began as—indeed were part of—oral epic tradition, and we have no reason to believe that the poems (as distinguished from the extant texts) were entirely cut off from that tradition for many centuries thereafter. "Homer" was known to all, apparently in many different forms, but the essence of the poems is likely to have remained quite intact within the shape-shifting of phraseology and narrative structure that was the very medium of their transmission.

Just as we cannot be sure that the Athenian text held pride of place as the koinê that apparently eluded the editorial signature of Aristarchus and his predecessors, so we have even less reason to believe that this same primordial text surfaced whole in the tenth-century Venetus Marcianus 454 and its brethren. In fact, from the very moment when (let us hypothesize) an amanuensis started to commit Homer's words to writing, there began an editing process that continues to the present day.^[18] Anyone concerned with mere intelligibility, not to mention aesthetic judgments, would have a hand in rendering the oral performance "acceptable" in grammatical terms at least, and this editing would still be necessary (if less frequent) for an oral-dictated text composed more slowly and carefully than a sung performance. From this beginning memorizers and rhapsodes alike, no matter how meticulous they

[17] We may compare the cases of the "literary" imitators of Serbo-Croatian oral epic tradition, such as Bishop Njegoš and Andrija Kacic-Miosic^[*] . See further Lord 1986a; Haymes 1980; Milerich 1978a, b.

[18] The inevitable editing associated with taking down a song from dictation may be glimpsed in (the amanuensis and gaslar ) Nikola Vujnovic's^[*] transcriptions of sung performances by the Parry-Lord guslari (e.g., text no. 6699 by Halil Bajgoric^[*] ; see chaps. 5 and 8 below). Lines and parts of lines are omitted, changed, and even added in the process. A second possible stage of editing is exemplified by the nineteenth-century collector Vuk Karadzic's^[*] occasional modification of texts collected by others before final publication (see Foley 1983b).

were in pursuit of perfect preservation, would continue to modify the received text and to shape it to their conception of what Homer must have said or meant. These processes, along with the important phenomenon of local traditions mentioned above, go a long way toward explaining both the multiplicity of editions of Homer (in addition to the hypothesized Panhellenic epitome) and the typical variation among papyri. The fact that these papyri start to die out after Aristarchus is perhaps attributable less to his reputation or influence as a textual scholar than to the simple phenomenon that by about the middle of the second century B.C. the last vestiges of a once-vigorous tradition are gone. The poetic tradition, having passed from aoidos to rhapsode and beyond, had to all appearances died out completely.

If we cannot be certain that the Byzantine texts look directly back to Homer, indeed if we cannot discover any verifiable connection even to Alexandria, are we then lost in a scholar's nightmare of editorial debris? Not, I propose, if we look at the evidence that actually exists and resist the impulse to impose stemmata and other generalized models that are the literary scholar's stock-in-trade. Consider for a moment the audiences of Homeric poetry in the primary oral phase; like audiences that attend the composition and recitation of oral epic narrative in other cultures, they "knew the story" of what was transpiring (Lord 1960, 99-123). Even the Panhellenic text, after all, had to have an audience, perhaps in that case a rather heterogeneous group accustomed to variations on what they were hearing, and this constituency heard the authoritative Homer against the echoes of the poems as they (no doubt quite differently and individually) knew them from prior experience. There was for them probably nothing especially sacred about this particular text except for its role in rhapsodic competitions, and even those singers and rhapsodes present at the Panathenaic Festival may well have returned to more parochial versions in other milieus.

Editions and papyri from the time of the Alexandrian scholars offer the same impossible problem for the literary critic seeking to confer a literary order on these materials, and present the same opportunity for those willing to seek another answer, one more faithful both to the available facts and to the period in which the texts were made and transmitted. The editions multiplied because the local traditions that began well before the sixth-century epitome continued to develop in their separate ways, dependent on rhapsodic activity and memorial transmission as well as on the written word. The papyri varied in ways that betray some sort of oral traditional transmission; even if we understand their production as exclusively a written process, we must make the minimal assumption that the writers knew and could use the oral traditional style, otherwise we cannot explain their characteristic variation.^[19]

In short, the journey of the two Homeric epic texts is most fruitfully and

[19] Cf. Lord 1986c on transitional texts; see also note 17 above.

faithfully understood as the journey of the Homeric poems . Whatever lacunae exist in the manuscript history (and there are many), we may be sure of this much: up to the time of Aristarchus, Homer's poems existed not in one exclusive version but rather in many versions of different but—from the perspective of tradition—equal authority. Held together by what Lord has so felicitously called a "tension of essences" and by an overall conservatism of language and thought possible only in a homeostatic idiom such as the archaic Greek poetic language, Homer's poems survived the centuries intact—from Homer's own texts through the Panhellenic standardization and on through the Alexandrian and Byzantine eras. Under these conditions, searching for a textual archetype must be as futile an exercise as searching for the archetypal form of a story. In the end it does not matter which manuscripts the Alexandrian scholars favored or exactly how the rhapsodes stitched in their own contributions; what matters is that the text we have is the result of a concatenation of processes, all of which served to "edit" the oral performance of Homer.

As a negative conclusion, we should thus be hesitant about attributing the term oral epic to the surviving Iliad and Odyssey without qualification. Given the uncertainties of transmission and the inevitable editing that Homer's songs underwent, it would be wiser to understand these texts as "oral-derived" in the sense advocated in chapter 1. Surely the poems' essence has survived the journey of the text, and we would be shortchanging Homeric art to deny its traditional nature; multiformity was an active part of text-making, and then of textual transmission, long after the advent of writing and Panathenaic epitomization of the text, and it remains a crucial aesthetic dimension of our Iliad and Odyssey , a dimension much in need of further investigation. At the same time, though, it would be at best romantic and at worst misleading to treat these same Homeric texts as if they were the completely unedited songs recited for dictation by a Yugoslav guslar . Especially when, as in the present volume, we seek answers to structural questions, we must be aware that while the Serbo-Croatian material is unambiguously oral and presents the welcome opportunity to measure a living oral tradition quantitatively, the Homeric poems as they have come down to us are oral-derived and cannot simply be equated with the epske pjesme for the purposes of comparison.^[20]

But positive conclusions are also possible and necessary. For example, the mélange of editions and papyri helps us to describe how Homer was transmitted and understood: in effect, his poems must have had an active audience versed in the oral traditional inheritance as late as the mid-second century B.C. Could it be that the poets of some of the medieval—and also oral-derived—

[20] As Mark Edwards (1986) cautions, we must be careful to consider variant readings when conducting statistical analyses designed to determine the possible orality of the Homeric poems. My purpose in these studies is not to attempt such a quantitative assessment but rather to clarify the tradition-dependent character of formulaic phraseology in each tradition.

epics addressed this same kind of audience?^[21] If the Alexandrians seem not to have affected the vulgate text in a major way, we may justifiably conclude that, owing in part to the absence of a book trade, the transmission of Homer was not in any important fashion a scholarly enterprise. Perhaps, like the verse romances of later medieval England, these poems were passed on as a kind of oral palimpsest, to be "erased" and re-"written" in accordance with traditional structure and within the limits of the multiform idiom. Most of all, we should recognize that Homer's poetry was itself a powerful enough medium and a sufficiently finely crafted instrument to assure its own preservation as it wandered through the bewildering historical maze of poets, rhapsodes, and editors. In their essentials, then, these are Homer's, and his tradition's, poems.

The Riddle of Cotton Vitellius A. XV.

If the Homeric texts that survive to our time provide only brief glimpses into a long and uncertain past, at least the evidence permits reasonable hypotheses to be drawn about their probable recording and the first standard texts, as well as about some of the prime historical influences which those or related texts must have undergone. Such is not the case with the Beowulf poetry, which comes to us in a single copy and whose date and provenance remain objects of lively debate among paleographers and students of the poem.^[22] We know very little even about the circumstances under which such a poem might have been collected, copied, and transmitted; with no real chronology for the verse of this period outside of a few termini a quo ,^[23] scholars have estimated the "date" of this version of the poem at anywhere between the mid-eighth and early eleventh centuries. There was no Alexandrian Library to (theoretically) hone the textual tradition, no set of scholia to provide fodder for readings or textual emendation, and no Byzantine scholarship to keep interest in the poem alive. Beowulf surfaces as a unique text in the middle of the sixteenth century, and is treated as a poem rather than an archaeological or paleographical curiosity only well into the twentieth century.^[24] All in all, the story of the text reads more like an Anglo-Saxon riddle than a Homeric epic journey.

[21] Especially considering the phenomenon of medieval diglossia; see Stock 1983, 30-87.

[22] For the older conventional arguments, see Klaeber 1950, cvii-cxx. For a sense of the variety of modern opinion, see Chase 1981a.

[23] "The Battle of Maldon," for example, celebrates a historical battle fought on August 11, 991.

[24] Notwithstanding, for example, Thomas Jefferson's introduction of Old English into the curriculum of the University of Virginia in 1825 (see Hauer 1983), Jefferson's interest being primarily in the language and law of the Anglo-Saxons. Only with J.R.R. Tolkien's justly famous essay "Beowulf : The Monsters and the Critics" (1936) did scholarly attention really turn toward the text as poetry.

The earliest identifiable owner of the manuscript, Laurence Nowell, dean of Lichfield and one of the first students of Anglo-Saxon, is also our earliest witness to the dated existence of the poem, and the date at which he signed his name was 1563. Shortly thereafter it came into the possession of Sir Robert Bruce Cotton (1571-1631), the noted antiquary whose library housed so many precious medieval manuscripts. The designation of the manuscript containing Beowulf as Cotton Vitellius A. xv. indicates its physical position in that library—in a case under the bust of the Roman emperor, on shelf A, fifteenth place. In 1731, one hundred years after Cotton's death, a fire at Ashburnham House, the location of the library at that time, destroyed much of the original collection. We are thus perhaps fortunate that the Beowulf text was merely singed at the top and sides. Over the years that followed, however, the originally relatively slight damage was worsened by handling of the now-brittle vellum leaves, and invaluable bits of the written record itself began to crumble away. Around 1870 authorities at the British Museum checked the process of deterioration by separating the manuscript book into single folios and installing a paper binding to protect the page edges.

Modern scholarship has benefited enormously from various transcripts made of parts of Cotton Vitellius A. xv. Even before the fire, Humphrey Wanley had recorded part of Beowulf , and Franciscus Junius had copied all of the accompanying fragmentary Judith that remained (Bodleian MS. Junius 105). But the most important transcripts for Beowulf scholarship are those made by the Icelander G. J. Thorkelin:^[25] two copies of the complete text written in 1787, about fifty years after the fire, when the manuscript was presumably in much better condition than it is now. The so-called A transcript was apparently made by a scribe whom Thorkelin hired for the purpose, and his errors and omissions reveal that the scribe was not well acquainted with the language in which he was working. Thorkelin himself wrote out the B transcript, however, and the professional knowledge of the philologist of Germanic languages shows through at every turn. Logically enough, this difference in preparation makes the A text usually more trustworthy than the more professional B text; since the amanuensis copied only what he could see with his untrained amateur's eye, he did not make educated guesses about emending crumbling words or filling lacunae.^[26] These two transcripts taken together provide an invaluable companion for editing Beowulf

From these materials—the unique text and its transcriptions—scholars have over the last two centuries established a poem epitomized in the edition of Friedrich Klaeber. We also have facsimiles by Zupitza ([1882] 1959) and Malone (1963) to use as evidence of the original manuscript contents, as well

[25] For the facsimile edition, see Malone 1963.

[26] This is not a universal principle, of course, since A could make errors of his own, especially with interpretation of minims. But when A omits a word or part-line altogether and B restores it in full, a careful editor will suspect the expert's intrusion.

as hundreds of proposals for repairing occasionally disturbed lexicon and syntax in the text.^[27] The emendations tend to enter the poem that we read and interpret relatively slowly; in fact, from time to time there have arisen among scholars movements to reverse the tide of emendation and to maintain original manuscript readings wherever possible.^[28] Behind questions of manuscript condition, transcription, and emendation, however, stands the seminal problem of the authority of the text as received. Unless we can gain some idea of exactly what Cotton Vitellius A. xv. offers us—and specifically with what authority it presents the poem—our decisions about consequent matters of editing and analysis will not be firmly grounded. And since the question of authority leads directly to dating, or early history, and to provenance, let us inquire first what can be discovered about the genesis of the text.

In an important sense, any attempt we might make to date the poem Beowulf would amount, in the words of David P. Henige (1964), to a "quest for a chimera." First and foremost, we are dealing with what is minimally an oral-derived text, a work that emerges from oral tradition along pathways we cannot now retrace and is intimately tied to that oral tradition, even if by the time our version of the poem was made the written word had become the sole textual medium.^[29] The formulaic phraseology and narrative multiformity typical of oral narrative composition comprise the very fabric of Beowulf ;^[30] although we may choose to insist on a literate editor or "author" for the surviving text of the poem, we cannot ignore its more than evident roots in an Anglo-Saxon tradition of tale-telling that arose and flourished without the aid of writing.^[31] This has been the most significant lesson of the excesses of the oral theory as propounded by Magoun ([1953] 1968), and equally of the intemperate reactions against that first statement. As described above in chapter 1, Beowulf need not answer all of the (tradition-dependent) criteria established by comparison with Serbo-Croatian and ancient Greek epic in order to be an oral traditional poem in a meaningful sense. Indeed, this greatest of surviving Old English poems may even have passed through one or more stages of composition or transmission that involved writing and still lay claim to oral traditional provenance. Once the critical dust has settled, the phraseology and narrative structure of the poem remain verifiably oral traditional, and we must confront that reality on its own terms.

[27] Especially noteworthy in this regard are the recent observations made by Kevin Kiernan (1981a) after a fresh examination of Cotton Vitellins A. xv.

[28] For a recent example, see Tripp 1983. In 1984 at the Modern Language Association, a panel entitled "Subtractive Rectification" pursued exactly this course.

[29] Cf. Renoir on oral structures even in later, written texts (e.g. 1976b, 1988); and we do not know that Beowulf is not yet closer to the oral tradition than any of the texts Renoir is examining.

[30] See the entries under "Old English" in Foley 1985; even Benson (1966) admits the formulaic structure of the poem.

[31] Consider the analyses of Niles (1983) and Opland (1980) on this point.

Studies in Homeric archaeology, for example, have illustrated how futile the quest for a uniform date for a traditional work must be: the assortment of weapons and armor in the Iliad , to offer a famous illustration, stems variously from Mycenaean through Dark Age times, with the individual elements in the armament never all current at any one time in history.^[32] Oral tradition does not recognize the anachronisms that result from our linear, non-recur-rent sense of history, any more than we would be content with tradition's customary interpretation of important societal events through pre-existent story-patterns.^[33] Thus it is that Beowulf's oral traditional character itself precludes our assigning a single date to the work; even had we been present at the (oral or written) composition of our Beowulf , we could not ascribe the poem to that very moment. The phraseology and narrative style that constitute the poetic medium, as well as the story of Beowulf that is its content, long predate that moment, and any attempt to compress the diachrony of the tale-telling tradition would amount to a falsification of the poem's structure, and therefore of its aesthetics.

Under such conditions it is only fair to admit necessary uncertainty about the "date of Beowulf " and to shift the inquiry to firmer ground, that is, to the nature of our text of Beowulf and to its witness in Cotton Vitellius A. xv. Allowing for the undeniably oral traditional roots of the poem, we could comfortably place this text of the poem anywhere in the eighth or ninth century, or perhaps a bit later. While over the years critical opinion has ranged all the way from the fourth to the eleventh centuries (Chase 1981a), moving the text earlier than the eighth or later than the early tenth century entails special problems. One faces either the lack of an appropriate historical or social context or, alternatively, the necessity of explaining too many linguistic or political anachronisms.^[34]

If we are to gauge the provenance of the Beowulf document, however, we must construct a reasonable hypothesis to cover the most significant ascertainable facts. I start with the minimal assumption of an oral-derived poem—

[32] See, e.g., Snodgrass 1974; and on the linguistic parallel, M. Parry 1932.

[33] Cf. Lord (1970, 27-28): "There is no reason to suppose that the patterns of ancient Greek epic tradition began with the Trojan War. They seem already to have been long in operation by the time of Homer, whenever one may place him in time. Historical events cannot give to a pattern the intensity and force needed for it to survive all the changes of tradition. These changes are not the corruption of time, but the constant reinterpretation of succeeding generations and societies. Not corruption, but constant renewal and revivification. The patterns must be suprahistorical to have such force."

[34] If recent scholarship in this area makes anything clear, it must be that we should keep an open mind on the date of the extant Beowulf text. In The Dating of Beowulf (Chase 1981b), contributors' opinions range from John C. Pope's argument for the eighth century through Kiernan's preference for the eleventh (see also Kiernan 1981a, together with the responses of, e.g., Boyle [1981] and Chase [1982]). On the dependability of linguistic criteria for dating, see Amos 1980.

that is, a text of demonstrably oral traditional character—and therefore with the tradition that must have preceded and given birth to our text of Beowulf . This minimal assumption, backed up by an appreciable amount of scholarship, does not demand positing an "oral poem," but it does entail recognizing that the creation of our text was necessarily more a process than an event. For a start, then, like most reviewers I must doubt the likelihood of Kevin Kiernan's (1981b) explanation of the poem as coeval with the manuscript; in fact, I see no reason to contend that the first text was coeval with the manuscript. Earlier scholarship assumed as a paleographical commonplace that Cotton Vitellius A. xv. represents a copy some generations removed from the "archetype," and I find no incontrovertible proof suggesting that we should desert that position. Most importantly with respect to the eleventh-century dating, if we do accept the argument that Kiernan proposes, we in effect cut the poem off from its oral traditional provenance, and that would be a serious error—just as serious an error, one might add, as would be tacit acceptance of Magoun's first formulation of oral traditional art, a formulation that underestimated the aesthetic dimension of the poem. We need a hypothesis that allows the work both its undeniable roots and at least the possibility of memorial or written "polish."^[35]

Placing the origin of our text in the eighth to ninth centuries (or perhaps the very early tenth) violates none of the paleographical, historical, or linguistic data per se and permits us to view the artifact in its oral traditional context. I would designate the beginning of the eighth century, in fact, as a terminus a quo for the introduction of the Beowulf story into the written medium. After that point the making of a written Beowulf , whether by oral dictation or as an autograph, would be a logical, reasonable enterprise. And although, given that manuscript production was exclusively a monastic enterprise and therefore not as subject to the influence of the political situation as some have suggested, I do not place as much emphasis on historical as on other aspects of available evidence, it can be said that the times were not uniformly inhospitable to the production of what has been called a Danish poem. Copies almost certainly intervened between that first text and Cotton Vitellius A. xv. (in a moment we shall consider the implications of the hypothetical transmission process). But if we settle on the eighth century as the earliest possible dating of this text of the poem, with the added proviso that it had to have emerged before two and one-half centuries had passed, we shall not be far from the truth. Most importantly, we shall not lose contact with the poem's lifeblood, its tradition.^[36]

[35] Compare what was said above in relation to the Homeric poems about the inevitability of editorial "polishing" during the process of transmission.

[36] Compare Pope's (1981, 195) remarks on this point, made as part of his argument for an eighth-century date: "I cannot imagine that the author himself did not have a long period of apprenticeship in the art of poetry before composing such a masterpiece. It would not be surprising if he had acquired most of his skill by oral composition. Certainly he had mastered a style well suited to oral delivery. As a pure Conjecture I would suggest that the oral poets of the seventh century were largely responsible for the growth of the style and the story, and for pointing the way, in courtly and sometimes monastic surroundings, for the remarkable adjustment of Christian beliefs and pagan story that Beowulf represents and the Icelandic poems and sagas (greatly to our present satisfaction) so conspicuously lack."

Manuscript Authority

Since we have even fewer clues to the transmission of Beowulf than to that of the Homeric poems, and since we are describing a work whose most essential style and content owe so much to a precedent and perhaps contemporary oral tradition, it will be impossible to build much of a case for manuscript authority on grounds acceptable to modern textual criticism. As we have seen, neither an original date for execution of the poem nor a chronicle of its transmission can be established. To these realities we must add the fact that, as Kenneth Sisam ([1946] 1953) has pointed out, the Old English poetic manuscripts generally give us little reason for confidence in their authority. He notes (p. 36) that

ample evidence from other sources confirms that copyists of Old English texts were not expected to reproduce their originals letter for letter, as they were when copying Latin and especially Biblical texts. Modernization of forms in the course of transmission was allowed and even required by the use for which Old English work were intended, and the practice was obviously dangerous for the wording.

Indeed, one cannot account for errors and changes by ascribing them to visual or aural mistakes, he continues, and "as compared with the variants in classical texts, they show a laxity in reproduction and an aimlessness in variation which are more in keeping with the oral transmission of verse" (p. 34).

An illustration of this last tendency (phrased perhaps too negatively, or at least literarily, by Sisam) may help to clarify the point. One of the few Old English poems to survive in two versions, "Soul and Body," exists in an incomplete Vercelli Book copy of 166 lines (usually called "I") and an ostensibly complete Exeter Book version of 121 lines ("II"). That these two texts are somehow related to a single poem is obvious: in addition to similar overall content, they share a large number of repeated or formulaically related lines.^[37] What is more, even the formulaic modifications occur at metrically predictable spots, with substitutions of words following prosodic rules.^[38] Further, most of the "unique" lines, those not shared by the two texts as we

[37] See Alison Jones's two articles on "Soul and Body" (1966, 1969), as well as the ASPR Exeter Book version, which marks repetitions and equivalents.

[38] We may also observe that the fourth stressed position in a line, the one stress that almost never participates in the alliterative pattern, is the most common locus for modification. Since it is not constrained by alliteration, this position can most easily tolerate the substitution of a synonym.

have them, echo against the larger referent of the Anglo-Saxon corpus as a whole, so that both texts are essentially traditional. What we seem to have in "Soul and Body" I and II, then, amounts to two versions of what was once a single poem, with each version having taken an idiosyncratic textual form best explained as a substantive traditional variant. While it is possible and even likely that these two texts diverged in part through individual histories of manuscript transmission, we should also admit the not exclusive possibility that memorial transmission in an oral context accounted for their formulaic resemblance.

Thus it is that the term manuscript authority really constitutes a misnomer in reference to the Old English documents. In addition to the many spelling changes and other typical scribal modifications, the editor and critic must consider the feature of formulaic, and perhaps memorial, transmission as well.^[39] Whether such a medium was used both for the shorter, more memorizable poems and for the longer works like Beowulf is a question we cannot answer with certainty, a problem in genre-dependence that we do not have the evidence to solve.^[40] But in assessing the authority of our received text of Beowulf , it is well to keep in mind that what Sisam has called "a laxity in reproduction and an aimlessness in variation" may be signs that the most important manuscript record for Old English poetry—like its Homeric counterpart essentially an oral palimpsest—was the memory.^[41]

Implications for Comparative Analysis

At first sight the Homeric and Old English documents may seem quite comparable, both preserving a dead-language poetry about whose history we have little hard information. But there the resemblance ends. The Beowulf manuscript is a unicum ; we cannot collate it with other texts, attempt stemmata, weigh precious bits of historical information, or even solve the minimal problem of the likely date and provenartec of the poem it encodes. Much more so than even the Homeric poems, the work we call Beowulf is severed from its origins and history by its very uniqueness and lack of ascertainable background. The text can be dated, I have argued, most easily to the period A.C. 700-950, but that assignment does us little good in the face of so many missing facts. Our poem may have descended through one or a dozen scribal reformulations, some more exacting (in our modern sense) than others; it may have experienced, in part or in whole, a memorial transmission that altered its original form considerably. The plain truth is that, unless dramatic new

[39] Compare the reconstructions of the memorial process in Fry 1974, 1981; Lord 1986c.

[40] See chapter 1 above and Foley 1983b.

[41] It is worth adding here that the tradition-dependent nature of Old English prosody would have allowed more variation than would the Homeric hexameter; see further chapters 4 and 6 below.

evidence emerges, we shall never know how far removed from its originative traditional composition this version stands.

As we did with the texts of the Iliad and the Odyssey , let us consider first the negative conclusions to be drawn from these remarks. First, since the Homeric and Old English manuscript records remain mysterious and idiosyncratic, no scholar interested in precise and sensible analysis can feel justified in quantitative comparison of these two remnants of two great traditions. The documents to be analyzed are vastly different not only in tradition (and therefore in language, prosody, phraseology, and even narrative structure) but also, and most basically, in their very physical make-up. If we cannot establish the primary oral nature of these texts, and if further we cannot certify that the items we are proposing to compare are truly comparable, then how is it possible to draw meaningful conclusions about, for example, formulaic density? The documents cannot be forced into a single category, any more than the languages or prosodies involved can be forced into absolute comparability.

But there are also positive conclusions to be drawn. Just as re-examination of the textual history of the Homeric poems indicates that the Iliad and Odyssey were but the epitome of an oral tradition that continued in some form well beyond their Panhellenic apotheosis, so the relative isolation of the Beowulf manuscript need not mean abandoning the search for poetic context. Although Beowulf , like the Iliad and Odyssey , must not without further proof be termed oral in the unambiguous sense that the Serbo-Croatian material collected by Parry and Lord is oral, neither can it be denied the designation oral-derived .^[42] Over the past thirty-five to forty years a great deal of evidence has accumulated indicating that Beowulf and other Old English poems show the oral traditional features of formulaic phraseology and thematic structure, and no one would argue today that the poem's roots are not planted firmly in an Anglo-Saxon oral tradition. Exactly how far (if at all) Beowulf is removed from primary oral tradition is a question we cannot settle, but neither should we "throw the baby out with the bathwater" by ignoring the poem's indisputable oral-derived character. Once that minimal and justifiable assumption is made and accepted, the way is clear to invoke the remainder of the Old English poetic canon as a context (with proper calibration) and to begin to describe the meaning of traditional elements in Beowulf by reference to their recurrence in other oral-derived poems.^[43] While we should be careful to acknowledge genre-dependence in this endeavor, and while the comparisons made within

[42] Although I prefer the connotations of Burton Raffel's term oral-connected (1986), since it emphasizes continuity rather than discontinuity, I maintain oral-derived in order to stress the difference between texts that are obviously and verifiably oral and those which exhibit oral traditional features but cannot be proven on other grounds to be primary oral texts.

[43] Or to expand the context beyond the Old English poetic canon to other Germanic and non-Germanic traditional poems; see, e.g., Renoir 1979b, 1981b, 1986.

the Old English canon cannot be (by definition) comparisons of parallel works from a primary oral tradition, this method of searching the oral-derived wordhoard for information about its constituents takes into account the more-than-literary relationships among surviving works. In short, what we lack in conventional literary-historical background about the document we call Beowulf can be in part replaced by the poem's most natural context—what has survived of its poetic tradition.

The Yugoslav Guslari and Their Tradition

When in 1928 Milman Parry underwent the soutenance associated with the presentation of his two theses on the traditional style of the Homeric poems, he was as yet unaware of the next step he would take in positing that a Homer who composed traditionally must also have composed orally.^[44] Parry attributed this conceptual leap to remarks made by his mentor, Antoine Meillet, during the defense and to the presence of Matija Murko, the Slovenian ethnographer who had been recording and studying the oral epic bards of the South Slavs for some years.^[45] Although Parry did not at that time recognize the importance of the lectures Murko was presenting in Paris (later to become Murko 1929), he observed in his field notes, "Cor^[*] Huso," that "it was the writings of Professor Murko more than those of any other which in the following years led me to the study of oral poetry in itself and to the heroic poems of the Southslavs" (SCHS 1:3). The advances documented in his classic articles on Homer and oral tradition (1930, 1932) had their point of origin in the 1928 soutenance and in the influence of Murko that became more significant as the years went on.

After a period of reading about the practice of oral poetry as described by Murko, Gesemann (1926), and Radloff (1885), and after realizing that his thinking on the subject was entirely theoretical, Parry determined to observe the phenomenon of oral composition at first hand.^[46] Always foremost in his mind, however, was the applicability of the research to ancient Greek epic; in his own words, "it was least of all for the material itself that I planned the study" (SCHS 1:3).^[47] Nonetheless, with a characteristic blend of imagination

[44] For a full account of Parry's work and its antecedents, see Foley 1988, chaps. 1-2.

[45] Murko's seminal but under-appreciated writings on oral epic in the South Slavic territories began as early as 1912; his major works were published in 1929 (the French monograph) and posthumously in 1951 (the two-volume magnum opus in Serbo-Croatian). My translation of these two latter works is in preparation (Foley forthcoming).

[46] See SCHS 1:3-20; Lord 1951b, Lord 1960, chap. 1; and Bynum 1974.

[47] He continues (SCHS 1:4): "In other words the study of Southslavic poetry was meant to provide me an exact knowledge of the characteristics of oral style, in the hope that when such characteristics were known exactly their presence or absence could definitely be ascertained in other poetries, and those many large and small ways in which the one oral poetry differed from written poetry for its understanding could be carried over to the Homeric poems."

and assiduousness, he carefully planned and carried out a program of collecting unequaled in any European tradition, before or since. The results of his efforts, including also expeditions undertaken in later years by Albert Lord and David Bynum, constitute the Parry Collection of Oral Literature at Harvard University, an archive of more than two thousand recorded songs, some of the best and most representative of which have been published in the series Serbo-Croatian Heroic Songs .

Parry's first trip in 1933 was chiefly an organizational mission, and he returned in 1934-35 to spend fifteen months recording and interviewing guslari . With his co-worker Albert Lord and native assistant Nikola Vujnovic^[*] , Parry visited six principal centers in his search for oral epic comparanda: Novi Pazar, material from which region is published in SCHS , vols. 1-2; Bijelo Polje, the homeland of Avdo Medjedovic^[*] , the finest guslar Parry encountered (SCHS , vols. 3-4, 6); Kolašin, in Montenegro; Gacko, represented in SCHS , vol. 14; Bihac^[*] (also represented in SCHS , vol. 14); and Stolac, the homeland of the singers whose epic narratives are in part the subject of the present volume.

Among the many songs and versions of songs to be found in the repertoires of guslari from these six centers, Parry soon discovered that a particular subgenre of epic would best suit his comparative work. As Lord points out (SCHS 1:16): "For his Homeric studies Parry found the songs of the Moslem population of Yugoslavia more significant than those of the Christian tradition, although it should be pointed out immediately that the singing tradition of both the Moslem Southslavs and their Christian brethren is the same." Parry preferred the Moslem songs because of their greater length, the result both of the Moslems having been the ruling class and of the continuous influence of the thirty-day feast of Ramadan, during which a month's worth of nightly entertainment in the coffeehouse was necessary. Under such conditions the singers of Moslem songs naturally developed much longer, more ornamental epic poetry, a poetry not unlike that of Homer.^[48] When we add that the Return Song (as described in chapter 1) occurs in both Christian and Moslem traditions but takes a much longer, more elaborate form in the latter, we can see where the most suitable comparand for the Homeric Odyssey lies—in the Return Song of the Moslem tradition.

In respect to the actual medium of performance, the songs collected by Parry and Lord are of three major types: sung texts, performed at customary speed (varying both by singer and within individual song-texts) to the accompaniment of the gusle; recited texts, performed without the instrument; and dictated texts, taken down in writing at a much slower pace by an amanuensis. Lord's (1953, 132) description of the special nature of this last category of song-text makes clear the major points of divergence:

[48] We may note that Parry was in effect observing a kind of genre-dependence by trying to suit the Serbo-Croatian comparand as closely as possible to the Homeric poems.

An oral poet who is asked to dictate a song for someone to write finds himself in an unusual and abnormal position. He is accustomed to composing rapidly to the accompaniment of a musical instrument which sets the rhythm and tempo of his performance. For the first time he is without this rhythmic assistance, and at the beginning he finds it difficult to make his lines. He can easily learn to do this, however, and he sets up a certain rhythm in his mind. He is also somewhat annoyed by having to wait between lines for the scribe to write. HIS mind moves ahead more rapidly than does the writer's pen. This technique he can also learn, particularly if the scribe is alert and helpful. The singer is accustomed to the stimulus of an audience, but again an intelligent scribe and a small group of onlookers can provide this stimulus.... The chief advantage to the singer of this manner of composition [oral dictation] is that it affords him time to think of his lines and of his song. His small audience is stable. This is an opportunity for the singer to show his best, not as a performer, but as a storyteller and poet.

In general, the oral-dictated text reveals fewer "errors" of all kinds: fewer "bad" lines (unmetrical or fragmentary verses), slips of the tongue (substitution of words such as near-homophones), nonsensical lines (whether or not these occasion a partial repetition immediately following), divergences in the narrative, and other blemishes in phrasing or plot structure. Such dictated texts arc often slightly longer than sung texts, since the guslar once accustomed to this medium can compose more carefully, without much of the pressure he otherwise feels in performance. These and other features have led Lord to posit that the Homeric poems as we have them must have originally been oral-dictated texts, and that Homer must have been familiar and comfortable with dictation. Whatever the case may be, I have selected both sung and dictated texts for the Return Song sample analyzed in these studies, in the hope that such an approach will pluralize our findings. The structural profiles of phraseology, narrative structure, and story-pattern, in other words, are not dependent on or limited by the medium of the songs examined.

One more issue must preface a closer look at the singers of Stolac and their songs. We need to recognize from the start a basic limitation on comparability in the case of the Serbo-Croatian witness. Whereas the ancient Greek and Old English texts are (minimally) oral-derived, the two histories of transmission diverge sharply; as noted above, various factors must calibrate any comparison we may wish to make. In the case of the Serbo-Croatian material, however, the problem of document comparability in a sense looms even larger: although our characterization of the work is not handicapped by lacunae in our information about memorial or manuscript transmission, the fact is that in Yugoslav oral epic no "text" exists until it is recorded—after which point its "polishing" by editors is either minimal or non-existent.^[49]

[49] This pertains, of course, to the Parry-Lord texts and not, for example, to the songs collected in the nineteenth century by Vuk Stefanovic^[*] Karadzic^[*] , which did undergo some editing (see Foley 1983b).

The Stolac songs, whether sung, recited, or oral-dictated, constitute oral tradition in its purest form, without the usual deflections inevitable in any transmission process. Yet at the same time, each "version" represents an equal performance variant of the song or "work" that itself remains forever untextualized.^[50] Before overall comparisons can be meaningful, they must be contextualized by a meticulous examination of each comparand on its own terms. In the case of the Serbo-Croatian material, this examination includes not only the points developed above but also the three Stolac singers and their repertoires of songs.

The Guslari of Stolac

Milman Parry and Albert Lord began their recording of oral traditional epic in the Stolac district of central Hercegovina, a region that in 1953-35 supported a strong and varied Return Song tradition. Songs by twenty-eight Stocani were gathered during the early expeditions, and later trips in the 1950s and 1960s by Lord and David Bynum added substantially to the sample of epic singing from this area. In later chapters we shall be looking both at the local tradition of the Stolac region in general and at three individual guslari who in particular help to comprise that tradition. Each has in his recorded repertoire a number of Return Songs, and it is principally to a selection of these works that we shall turn for evidence on phraseological and narrative structure in an unambiguously oral epic tradition.

Ibrahim Basic^[*]

Although he spent the greater part of his life in the Stolac area, Ibro Basic^[*] was born in Vranjevici^[*] in the district of Mostar. At the age of fourteen he left his father's household to become a servant to Ahmet-efendija, a wealthy landowner from Opijac: in Dubrava, with whom he remained for about ten years. He then entered the service of Salihagha Behmen on the occasion of his new master's wedding. After three or four years he himself married and lived for a time with his bride, Djula Dzanko, in her village of Osanjica^[*] . Subsequently they left the village and Ibro found work as an attendant in a coffeehouse (kafana ).^[51] During this period Djula became quite ill and her brother took her home to live with him in the hope of improving her health.

[50] See further Foley 1986a.

[51] Although he does not tell Nikola anything further about the coffeehouse experience, it seems safe to speculate that during this time Ibro had an opportunity to hear many, singers perform. He may well have learned from them and perhaps even performed there himself. A fragmentary description from Parry text no. 291b and an elaborate episode from no. 6598 indicate that he had at various times been paid for singing publicly. I employ the standard Parry Collection notation in citing the texts by number (for a digest of material collected through 1951, see SCHS 1:21-45); the italicized numbers indicate texts that were recorded from actual oral performance, the Roman numbers those taken down in writing from dictation.

Ibro soon followed her back to Osanjica^[*] but became bedridden himself with a tumor in his leg.^[52] He remained in the village for some eleven years after his recovery, living with a Ratkusic^[*] family. Djula died shortly thereafter, and a childless remarriage lasted only four years. By 1934-35, when Ibro was about seventy years of age, one son had moved to Belgrade and a second was with his father working at their modest woodcutting trade.

Ibro began to learn to sing at the age of eight or ten,^[53] first from his father, a skilled guslar who was part of the local tradition in which a singer characteristically performed in villages near his own. While in the service of Ahmet-efendija Ibro also often traveled to nearby villages and heard guslari sing. In those years, he recalls, he was able to hear most songs only once and remember them well enough to perform them. Along with his father, Ibro also encountered five more singers of varying age and reputation in his early years. From Sule Tabakovic^[*] he learned Djerdelez Alija and the Ban of Karlovo (Parry nos. 291, 291a , 6596);^[54] from Ibro Coric^[*] , Grga Antunic^[*] Attacks Raduc (no. 6692 ); from Osman Marijic^[*] , Hrnjicic^[*] Mujo and the Ban of Karlovo (no. 645) and The Wedding of Smailagic^[*] Meho (no. 12491);^[55] and from Selim Basic^[*] and Selim's father, Alagic^[*] Alija and Velagic^[*] Selim (nos. 291b , 1283, 6597 ).^[56]

In addition, we hear in conversation no. 6598 of a semi-legendary guslar of a previous generation, a certain Isak of Rotimlja by name, who was summoned for only the grandest occasions and was always splendidly rewarded for his performances.^[57] Isak's legendary accomplishments include invitations to play before beys and pashas, a paradigmatic triumph in a song contest with a lesser singer named Gacanin, and requests for his presence at weddings of all religious denominations. Ibro seems never to have actually met Isak, and the complete lack of historical context and personal contact as well as the stories of his remarkable feats lead one to believe that this greatest of singers belongs to the same world of folktale inhabited, for example, by the Anglo-Saxon scop

[52] This illness is a very "heroic" one, reminiscent in its hyperbole of Salih Ugljanin's account of how to behead a man (SCHS 1:63-64, 2:5).

[53] On the special gift of youth in learning to sing, Ibro says (Parry no. 6598 ): "A child of four years naturally remembers more than if he were twenty. The child is a great wonder; he can learn and retain it in his head. It's in the nature of a child." Cf. Lord's description, "Singers: Performance and Training" (1960, 13-29).

[54] In text no. 291b he credits Selim Basic^[*] , his father, as the source of this song; this is the only contradiction about sources, however.

[55] This last text, recited without the gusle and imbedded in the very lengthy conversation no. 6598 , demonstrates the important pacing function of the instrument. Hypersyllabic lines abound; compare SCHS 1:107-16 and 117-26.

[56] Ibro explains that he heard the last song first from Selim himself and later from Selira's ninety-year-old father, who claimed to have been a "standard-bearer" (bajraktar ).

[57] In text no. 6598 Ibro says that "Isak was the best singer in Hercegovina" and that he was always paid because, unlike all the other guslari , "it [singing] was his profession" ("to je njemu zanat bijo").

Widsith.^[58] At any rate, Isak seems to serve for Basic^[*] , and for other singers, as a folk anthropomorphization of the epic singing tradition that they also embody.

Other sections of Ibro's conversations with Parry's native assistant Nikola make it clear that the guslar considered his songs "true," and yet that he also had a concept of what has often been called "ornamentation." For example, when asked whether the story had truth (istina ) in it, Ibro says: "All is true, I believe, yes, even though some things are added (as you know)^[59] to make it more fitting [zgodnije ]; but there were all sorts of things then—there were heroes (and in yet earlier times there was a great number of them), and there were horses and swords and all. It was not then as it is today" (no. 6598 ). In other words, in singing he is recalling or re-creating a heroic age when the events that make up his and others' songs actually took place, and although that time is far removed from the present day, he believes in its reality. In order to portray that age and those events in the most "fitting" way possible, he and others "add" the stateliness and grandeur of the epic tradition. This concept of the truth and its embodiment echoes Lord's (1970, 28) memorable observation on the relationship between historical truth as we know it and its representation in oral epic tradition; he notes that the stories are primary because "their matrix is myth and not history; for when history does have an influence on the stories it is, at first at least, history that is changed, not the stories."

Ibro also gives us an insight into the singer's craft in his response to Nikola's probing about the "accuracy" of repeated performances of a song—whether he composes and re-composes "word for word" each time—and about the very nature of a "word" (rijec or rec ) in a song. Consider this excerpt from conversation no. 6598 :

Nikola : What is, let's say, a rijec in a song? Give me a rijec from a song.

Ibro : Here's one, let's say, this is a rijec : "Podranijo od Kladuše Mujo, / Na vrh tanke nacinjene kule" ["Mujo of Kladusha arose early, / At the top of the slender, well-made tower"].

Nikola : But these are lines.

Ibro : Well yes, but that's how it is with us; it's otherwise with you, but with us that's how it's said.... But here's one, let's say, that is [a rijtc ]: "Podranijo od Kladuše Mujo," let's say, a rijec for "podranijo" ["arose early"]; "Prije zore i ogranka sunce" ["Before dawn and the sun's rising"], that's a rijec for "podranijo, uranijo, podranijo," so.

Apart from clearing up the confusion over what guslari really mean when they claim to perform songs "word for word" (rijec za rijec ) each time,^[60] these observations illustrate how the singer conceives of the units in his songs. Simply

[58] See further Foley 1978b.

[59] Nikola was also a practicing singer, and Ibro here assumes his knowledge of the role of ornamentation in the making of a song.

[60] Compare the very different and limited perspective of G. S. Kirk (1962, 100) on the South Slavic guslari : "These particular poets pay lip-service to the ideal of complete accuracy in reproduction, and are under the impression that they come very close to it. They are, in fact, far too optimistic; and their very confidence and lack of self-criticism prevents them from trying to achieve a higher standard of accuracy, which certainly lies within their power. The truth remains that even within their simple and unsophisticated oral tradition, with its incomplete formular technique, poems—not merely 'substance' or 'technique'—are transmitted, though with some variation and contamination."

put, his words are poetic lines, units that epitomize what Parry called an "essential idea" and which are governed. by the metrical structure of the tradition. Ibro is telling us that the idea "arose early" equals, or can be expressed by, a decasyllabic expression in a song, and implicitly that lines, groups of lines, and perhaps metrical segments of lines are his rijeci . Units of typographic description, demarcated by white spaces that serve as silent reading mnemonics, have no place in his "emic" or "ethnic" grammar of poetic diction. We would do well to keep his quite sophisticated observations in mind as we embark on a study of Serbo-Croatian oral epic phraseology.^[61]

Halil Bajgoric^[*] :

The second of our three gudari from the Stolac region, Halil Bajgoric^[*] , lived in Dabrica, a village so remote that he had to travel for three to four hours to reach the town in which Parry and Lord were recording and interviewing singers from that region. At the first encounter in 1934, Bajgoric^[*] was only thirty-seven years of age, unusually young for a guslar of his accomplishment (Lord was to record more songs from him seventeen years later). Of his personal history this singer says less than most. He relates stories of beys and their descendants in Dabrica and an engaging account of his grandfather's being tricked by a bey into forfeiture of his land over an unpaid loan. This unhappy turn of events in Montenegro led the old man to settle in the village of Blagoja, where both Halil and his father were born. Such family history follows story-patterns well known in the epic songs, such as escape from captivity, and the singer ornaments them with details from the epic tradition, such as lengthy catalogs of items or people. Throughout his conversation with Nikola, Halil time and again summons this kind of traditional idiom to tell legendary stories from his family's and district's past.

Bajgoric^[*] began to learn to sing as a young boy, first in emulation of his father and specifically in order to join him in performing in the coffeehouse or at a wedding or other celebration.^[62] Although his father was, like himself,

[61] See further chapter 5. On the distinctions between "emic" or "ethnic" categories on the one hand and "etic" or "analytic" categories on the other, see Ben-Amos 1969 and Dundes 1962. In the case of the guslar's description of his compositional units, however, we should take the "emic" perspective especially seriously, since these same units are responsible for the compartmentalization and articulation of traditional meaning. See Foley 1986a.

[62] It is typical of Bajgoric^[*] that he speaks somewhat romantically of the first stages of this process (no. 6698 ): "So I stole the gusle from my father. I went into another room, and when he went to sleep I would sing a little."

a farm laborer and not a professional in any sense, he enjoyed a reputation for being the finest guslar "in three districts." Bajgoric^[*] makes no secret of the fact that he too is held in high esteem by his fellows, nor that he is customarily rewarded generously (albeit in kind) for his performances. While most of the songs in his repertoire came directly from listening to other singers, Halil admits that he did learn one, Tsar Scepan's^[*] Wedding , from a songbook by having it read to him. In fact, he might well have learned more from that same source had not the process been so time-consuming and the selection, in his view, so parochial.^[63] As it stands, then, he designates his father and one Ilija Braduric^[*] , the latter of whom he says with some approbation was literate and could therefore draw his material directly from songbooks, as the sources for his own songs.

In addition, Bajgoric^[*] was among many guslari who spoke of a legendary singer from the past, an idealized figure whom they characteristically distinguished from the men who actually taught them their craft. Gifted with special talents and sought after in many quarters, Hasan Coso^[*] —as Halil called this figure—was a singer of wide experience: "He traveled everywhere throughout the world. And he lived for one hundred twenty years." Despite Halil's efforts to locate Hasan in real time and space (he is said, for example, to have spent most of his life in Dabrica), the nature of the biography reveals that this best of singers, like Basic's^[*] Isak, was more a symbol than a fact: "My God, he died very long ago; from what they say it was probably seventy years ago. He was not even my father's father."^[64] While he claims that his father learned to sing from Hasan, Halil himself denies any personal contact with the man. And although he indicates that his father and Hasan lived no more than a kilometer apart, it becomes apparent that this greatest of singers was unique in the community that surrounded him: "It is said that he could still jump twelve paces a half-year before he died. People say he neither dug nor plowed, nor did he ride a horse, but he always carried a rifle and some goods on a beast of burden, and thus traveling lightly he enjoyed himself and played the gusle. " From these and other indications, it is clear that Bajgoric^[*] effectively understood Hasan as a kind of personification of the epic tradition—an anthropomorphic focus for the stories and wisdom of oral epic not unlike Ibro Basic's^[*] Isak.

The songs passed down from such a paragon are of course "true," Bajgoric^[*] assures Nikola, and he is prepared to gloss any item or feature of the story as he related it. One of these textual footnotes, most of which concern names or minor events in a given narrative, offers a glimpse of the guslar's own

[63] The book in question, which, while never identified, sounds a great deal like some version of the poems collected and published by Vuk Karadzic^[*] , contained only Serbian songs and therefore no Moslem songs, particularly no licke pjesme ("songs of the Lika").

[64] Further probing established this last expression as an idiomatic way of denoting relative age rather than as an assertion of kin relation.

attitude toward traditional diction. On being asked for the origin of the ubiquitous toponym Markovac, Bajgoric^[*] replies that the term designates the mountain village in which the great Serbian hero Marko Kraljevic^[*] was nursed. This assignment of the toponym naturally means that the village bears a name unsurpassed for its honorific and heroic import. But then Nikola confronts Halil with a problem: if this interpretation is correct, how does he explain the verse formula "Pa eto ga niz Markovac kleti " ("There he is below accursed Markovac")? His suggestion of Markovac as the revered cradle of Serbia's most significant hero is at odds with a line he has himself sung; Bajgoric^[*] therefore falls back on the explanation that in a song "mora da se rekne" ("it has to be said [like that]"). As Lord and others have argued, the tradition speaks diachronically in the synchronic performance of the individual guslar , and that person need not—and often does not—consciously analyze and understand the nature of the idiom he is employing by right of succession. It is enough for Bajgoric^[*] to know that "Pa eto ga niz Markovac kleti" constitutes a rec , a traditionally defined and indivisible "word," which he can draw from his compositional lexicon in the making of his song.

Like Ibro Basic^[*] , Halil conceives of the action described in his songs as having actually taken place much earlier, in some sort of "Golden Age," and he offers the observation that most of the heroes involved, such as Aliagha Stocevic^[*] and Mustajbeg of the Lika, lived at approximately the same unspecified period in history. When he sings of these and other heroes, moreover, they are subject to the same traditional ornamentation used by Basic^[*] . At times this compositional flexibility—that is, the multiformity that is the lifeblood of the oral epic tradition—leads to what we might term an outright error or omission, the kind of narrative blemish that has so often influenced Homeric scholars to picture the great poet "nodding." A case in point is Nikola's calling to Halil's attention the omission of the hero's preparations for travel from a second version of a particular song.

Nikola : Yesterday you sang this song, but today when you sang it you skipped over one whole section that is usually sung—for example, when Marko readies himself and his horse for travel.... So did you shorten [the song]?

Halil : It's possible that I shortened some "words" [reci ].

Nikola : Yes, yes. Did you do it intentionally?

Halil : No, I didn't; it was only an oversight [in the] heat of performance.^[65]

On closer inspection it becomes apparent that the "words" that the guslar "skipped over" amounted to the usually paired themes of the hero arming for battle and readying his horse. We may draw two conclusions from these observations. First, Bajgoric^[*] understands the omitted section of the song in terms of reci or "words," the same unit described by Basic^[*] in reference to a

[65] Bajgoric^[*] speaks here of a velika vatra (literally, "great fire") that temporarily addled his thinking and caused him to stray from what he considered the proper story line.

line or group of lines in a song. Second, he not only is aware of the possibility of ornamentation by means of multiforms but also falls victim to the other side of the same process; flexibility of narrative structure means that occasionally a singer will delete one or more elements from what is customarily a sequence. As we shall see again in reference to the works of the next singer, this deletion may take place without the guslar's conscious notice. It is the price one pays for a generative compositional idiom that tolerates substitution and variation within incremental limits.^[66]

Mujo Kukuruzovic^[*]

The third of the Stolac guslari whose recorded works we shall be examining, Mujo Kukuruzovic^[*] , lived originally in the village of Grbavica in the district of Mostar and was forty-three at the time of the initial Parry-Lord fieldwork. Like both Basic^[*] and Bajgoric^[*] , he could neither read nor write, and like Bajgoric^[*] he earned his living farming land he had inherited from his father. Although he seems to have learned the craft of singing in the customary way, Mujo was unique among this group in the variety of sources he claims for his songs: he attributes his repertoire of thirty-eight different items to a long list of singers, eleven in all, none of whom was apparently related to him.^[67] He also classified each of these singers into one of two categories, the narodni (or "folk") guslar who was not for hire but could be rewarded for his services with food and drink, and the coffeehouse singer (guslar u kafani ) who was paid by the proprietor for his performances according to a pre-arranged agreement. We find both kinds of singers among his putative tutors, although Mujo's own preference is for the less commercial and less professional narodni guslar .

Mujo's conversation with Nikola proves him a lively informant, not least because of his willingness to offer an opinion on nearly every aspect of the singer's métier. While we cannot accept everything he says without critical evaluation, especially since like other guslari he tends to spice answers to questions about his personal history with the heroic details of his oral epic tradition,^[68] his comments on the art and process of singing are without doubt the most perceptive of those made by any of the three Stocani. For example, he knows of songbooks but says he does not use them because as a rule they contain only the newer songs, not the longer Moslem epics to which he feels a special affinity. And he is never slow to affirm his own ability to master the songs he has heard from others or had read to him. During discussion of the

[66] On the meaning of such variation and its context, see Foley 1984b.

[67] Note, however, that the two accounts of attribution, one in the repertoire (no. 1287) and the other in the conversation (no. 6619 ), at points differ on which singer is to be credited as the source for a given song.

[68] He claims, for example, to have been imprisoned in Hasan Pasha Tiro's tower! Hasan Pasha Tim is a character from the epic, most prominent in The Wedding of Smailagic^[*] Meho , (see SCHS , vols. 3-4).

Zenidba bega Ljubovica^[*] (The Wedding of Bey Ljubovic^[*] ), for instance, he tells Nikola how easily he can pick up songs to enlarge his repertoire:

Mujo : So now, brother, you go ahead and find some song I don't know; then, brother, read it to me and give me the gusle ; if I get confused, I'll give you a finger off my hand.

Nikola : And you'll sing the whole song this way?

Mujo : I'll repeat every, single rec .

Nikola : Well, that's a wonder to me.

Mujo : It'll be so, I guarantee it.

We might view this exchange as mere bravado on Kukuruzovic's^[*] part, and perhaps it is just that, at least on one level;^[69] but later on in the conversation he glosses the term rec memorably by echoing Ibro Basic's^[*] and Halil Bajgoric's^[*] concept of the traditional "word." After defining a rec as "an utterance [besjeda ] which comes after another as that [former] one passes forth, as it goes, as it falls out all in order—from beginning to end, as it is ordered and placed and so on," he prescribes one or more lines as the unit rec :

Nikola : Let's consider this: "Vino-pije-licki-Mustajbeze" ["Mustajbeg of the Lika was drinking wine"]. Is this a single rec?

Mujo : Yes.

Nikola : But how? It can't be one : "Vino-pije-licki-Mustajbeze."

Mujo : In writing it can't be one.

Nikola : There are four reci here.

Mujo : It can't be one in writing. But here, let's say we're at my house and I pick up the gusle —"Pije vino licki Mustajbeze"—that's a single rec on the gusle for me.^[70]

Nikola : And the second rec?

Mujo : And the second rec —"Na Ribniku u pjanoj mehani" ["At Ribnik in a drinking tavern"]—there.

Nikola : And the third rec?

Mujo : Eh, here it is: "Oko njega trides' agalara, / Sve je sijo jaran do jarana" ["Around him thirty chieftains, / All the comrades beamed at one another"].

In fact, not only does Kukuruzovic^[*] sense what Lord has called the "adding style" of oral epic and manage to communicate to Nikola a rather sophisticated notion of the functional kernel of traditional composition, but he is also able to conceive of the separateness of the four units that we, outside of the traditional idiom, would call "words." Although Mujo's comments on the sources of his songs may be contradictory and his estimate of his own ability as a guslar a bit inflated, his analysis of the units of compositional meaning—the reci that make up his song—is extremely enlightening. Quite naturally, with-

[69] See the discussion of his revealing departure from the story line in chapter 10 below.

[70] Notice that Kukuruzovic^[*] has metathesized the words vino and pije in this second mention of the line. Such flexibility is just one characteristic of the compositional "word" employed and deployed by the guslar . See further chapter 5.

out the scholarly precision possible only for an outsider but from a perspective available only to a traditional singer, he is very close to describing formulaic composition.

Kukuruzovic^[*] is also aware of the possibility for okitinje , or ornamentation, of the story as it is performed. In response to Nikola's questions about exactly what such ornamentation consists of, he is less than clear, but he does comment, for example, that he can sing a song that will prove "more harmonious" or "more integrated" than those customarily found in songbooks. He considers the stories basically true, although he retains a pragmatic attitude toward the relative "truth" of versions by various singers: "You know, I count as true that which I've more often heard." As we shall see in chapter 10, the question of multiformity at the narrative level—of ornamentation and different versions of a song—looms large with Kukuruzovic^[*] . For along with generic song-types of (a) Return and (b) Alliance with the enemy, he includes in his repertoire what we shah call a song amalgam , that is, a third type that is an additive combination of (a) and (b). We shall be able to watch the progress of traditional tale combination by tracing the outlines of the two constituent songs in the larger song and by observing the kind of connective tissue provided by the tradition to bind the two into a greater whole. What is more, the suture had not entirely healed at the moment that Mujo's repertoire was recorded, and the "narrative inconsistencies" (this time of major proportions) in one of the composite texts will allow us to follow the logic of guslar and tradition in the act of song-making.

The Parry Collection Texts Used for Comparison

For phraseological analysis I have chosen a unified group of five Moslem epic songs from the singers of Stolac, as summarized in table 1.^[71] These selections, made from a total of twenty-one songs I have edited from the singers of this region, reflect a mixture of media (three dictated and two sung texts) but a near unanimity of subgenre (four Return Songs and one Wedding Song). The sample thus includes both sung and dictated material from two singers and represents a unified local tradition. Taken together, the five songs total 7,287 decasyllabic lines, an extensive textual basis on which to found an analysis of oral traditional phraseology first within the local Stolac tradition and, by example, elsewhere within the Serbo-Croatian epic tradition.

For the major part of the thematic analysis, I have selected a group of eight Stolac songs in the Return subgenre that were sung or dictated by Basic^[*]

[71] The titles given here were assigned by Parry and Lord or, if they provided none, by me. Strictly speaking, titles are quite irrelevant to the guslar , who remembers a song in relation to its major figures and central action. The code numbers assigned reflect the protocol of the concordance and will be used in chapter 5 on traditional phraseology in Serbo-Croatian. See further chapter 5, note 23.

or Kukuruzovic^[*] (table 2). Near the end of chapter 8 comparisons are also drawn to a single song by Salih Ugljanin of Novi Pazar (SGHS 2 , no. 4). Once again we have both sung and dictated texts by each of two guslari from the same local tradition, all of them in the same subgenre, with the total sample amounting to 11,363 lines.^[72] This large and multi-layered inventory of songs and versions will allow exploration of individual, local, and pantraditional features of the narrative theme.

Before these analyses can begin, however, we must turn to an examination of the prosodies that exist in symbiosis with the ancient Greek, Old English, and Serbo-Croatian epic phraseologies, and which thus ultimately figure in the verbal expression of narrative patterns.

[72] Line-count per se is of course not as important a consideration in thematic as in phraseological analysis, but it does afford some idea of the extent of the sample used for thematic investigation. See further chapter 8.