rofessor joseph J. duggan's computer-aided investigation of
"formula" density in Old French epic poetry, the results of which
were published in 1973 in his much debated book, The Song of
Roland: Formulaic Style and Poetic Craft,1was the first extensive quanti-
tative application of the theory of oral composition developed by Milman
Parry and Albert B. Lord2 to entire poems and is in many respects a major
contribution to the field of ancient and medieval epic criticism. However,
his conclusion that the Chanson de Roland, like the other chansons de
geste he studied, is an orally composed poem3 has not, by any means,
brought general consensus among Old French scholars. One of the main
points of criticism justly raised against Duggan's study is that his criteria
for distinguishing between oral style and written style are not based on
extensive observation of "formulaic" density in authentically oral and
written texts.4 To date, indeed, relatively little has been done concerning
'Joseph J. Duggan, The Song of Roland: Formulaic Style and Poetic Craft (Berkeley,
Los Angeles; London: University of California Press, 1973).
2Milman Parry. "Studies in the Epic Technique of Oral Verse-Making: I. Homer and
Homeric Style." Harvard Studies in Classical Philology 41 (1930), pp. 73-147; Albert B. Lord,
The Singer of Tales (Cambridge: Harvard University Press. 1960; reprint ed.. New York:
Atheneum, 1974); idem, "Homer as Oral Poet." Harvard Studies in Classical Philology, 72
(1968), pp. 1-46. For accounts of the Parry-Lord theory, particularly with regard to quantita-
tive analysis, see. for example. Duggan, pp. 16-18; John S. Miletich, "The Quest for the
'Formula': A Comparative Reappraisal," Modern Philology, 74 (November 1976), pp.
111-123. See also Edward R. Haymes, A Bibliography of Studies Relating to Parry's and
Lord's Oral Theory (Cambridge: Harvard University Printing Office, 1973) and review of the
latter with additions by Samuel G. Armistead, MLN [Modern Language Notes], 90 (March
1975), pp. 296-299. John Miles Foley is now preparing for publication an updated bibliogra-
phy on the oral theory.
3This, of course, does not include Buevon de Conmarchis, which is known to be a writ-
ten composition and which Duggan classifies as a romance (see Duggan, p. 26).
4 Olifant / Vol 9, Nos. 1&2/Fall & Winter 1981
the establishment of such criteria from the standpoint of "formula" analy-
sis. Although oral-formulaic studies have greatly enhanced our under-
standing of the art of poems such as the Roland, they appear to have failed
to provide a satisfactory answer to the problem of their mode of composi-
tion, and this essentially for two reasons: (1) they lack extensive statistical
and descriptive data concerning oral material and deliberate imitations of
oral texts to which the poems under investigation can be compared;5 (2)
they have fallen short of applying a consistent method to various texts
4For a criticism of Duggan's criteria for the differentiation between oral and written
styles in narrative poetry, see, for example, John S. Miletich, review of The Song of Roland:
Formulaic Style and Poetic Craft, by Joseph J. Duggan, Modern Philology, 73 (November
1975), pp. 180-181; idem, "The Quest." pp. 119-121, 122-123; idem, "Études formulaires et
épopée européenne," in Charlemagne et l'épopée romane: Actes du VIIe Congrès Interna-
tional de la Société Rencesvals, Liège, 28 août - 4 septembre 1976, Bibliothèque de la Faculté
de Philosophie et Lettres de l'Université de Liège, 225, Les Congrès et Colloques de l'Univer-
sité de Liège, 76, 2 vols. (Paris: Société d'Édition "Les Belles Lettres," 1978), Vol. 2, pp.
428-429; idem. "Stilističke razlike između usmene i pisane književnosti: savremeni metodolo-
ški pristupi" ["The Stylistic Differentiation of Oral and Written Literature: Current Method-
ologies"], in Naučni sastanak slavista u V ukove dane: referati i saopštenja, Beograd - Priština
- Tršic, 13-19. IX 1976. (Beograd: MSC. 1977). 6/2, pp. 121-122 (I acknowledge the assistance
of John S. Miletich for the information in Serbo-Croatian}; Rudy S. Spraycar, "La Chanson
de Roland: An Oral Poem?" Olifant, 4 (October 1976), pp. 63-74. For a criticism of Spraycar's
study, see John S, Miletich's comments in "1976 Annual Meeting of the Société Rencesvals,
American-Canadian Branch: Proceedings." Olifant, 4 (March 1977), p. 171; William Calin.
"Littérature médiévale et hypothèse orale: une divergence de méthode et de philosophie."
Olifant,8 (Spring 1981), pp. 278-279.
5A comparative formulary analysis of two entire poems, a twentieth-century oral poem
from the Milman Parry Collection and a nineteenth-century poem written by Petar II Petro-
vić Njegoš was completed by Edward R. Haymes, who concluded that his results do not allow
us to distinguish clearly between "oral" and "written" material on the basis of "formula" den-
sity. Haymes found that the oral text contained 34.8 percent "formula" and the written text as
much as 29.6 percent: Edward R. Haymes, "Formulaic Density and Bishop Njegoš," Com-
parative Literature, 32 (Fall 1980), pp. 390-401. Duggan himself recognizes that we know very
little about the formulaic density of the Yugoslav poems, when he attempts to minimize the
importance of Haymes's conclusion by saying that more analyses would be necessary in order
for those results to acquire some significance with regard to a generalizing statement on for-
mulaic content and mode of composition: Joseph J. Duggan, "Le Mode de composition des
chansons de geste: Analyse statistique, jugement esthétique, modèles de transmission." Oli-
fant, 8 (Spring 1981), pp. 289-290, n. 1. Other as yet unpublished formulary analyses of entire
oral songs include Kenneth Goldman's doctoral dissertation (Albert B. Lord, "Perspectives
on Recent Work on Oral Literature," Forum for Modern Language Studies, 10 [July 1974], p.
189) and David E. Bynum's recent work (in a paper delivered at the MLA session entitled
"Yugoslav Oral Literature: Style, Structure, and Aesthetics" [San Francisco, 29 December
1979], Bynum announced that he had undertaken an extensive analysis of authentically oral
Carton / Oral-Traditional Style in the Roland 5
belonging to different literary and linguistic traditions, thus making it dif-
ficult to compare accurately the results which may be obtained for works
such as the Yugoslav oral songs on one hand and the chansons de geste on
the other.6 The validity of the criticism directed against Duggan's conclu-
sion on the orality of the Chanson de Roland is further corroborated by my
own application to the Old French text of another comparative and quan-
titative method of stylistic analysis, which is based on the observation of
significant differences in narrative mode between oral-traditional and
written poetry. This method was developed by Professor John S. Miletich
and does not deal with the "formula" but with the study of "elaborate
style," i.e., a delay in the flow of the narrative line through the use of cer-
tain kinds of repetitions.7 The results I have obtained show that the narra-
6On differences in "formula" analyses, see Miletich, "The Quest." pp. 114-120; Margaret
Chaplin. "Oral-Formulaic Style in the Epic a progress report." in Medieval Hispanic Studies
Presented to Rita Hamilton, ed. A. D. Deyermond (London: Tamesis. 1976), pp. 13-14; John
Steven Geary, Formulaic Dictionin the"Poema de Fernán González" and the"Mocedades de
Rodrigo": A Computer-Aided Analysis, Studia Humanitatis (Potomac, Maryland: Porrúa,
1980). pp. 6-11. As suggested by John Miles Foley, it may be necessary to consider the defini-
tion of the "formula" in view of the tradition in which it is used (John Miles Foley, "Beowulf
and Traditional Narrative Song: The Potential and Limits of Comparison," in Old English
Literaturein Context: Ten Essays, ed. John D. Niles [Cambridge, England and Totowa, New
Jersey: Boydell & Brewer, Rowman & Littlefield, 1980], pp. 117-122, 173-176), but, in doing
so, one must be aware of the significance that a given modification may have with regard to
eventual comparisons with material in other languages and traditions (Miletich, "The
Quest," p. 116).
7John S. Miletich. "Repetitive Sequences and their Effect on Narrative Style in Spanish
and South Slavic Traditional Narrative Poetry" (also appearing in bibliographies as "The
Romancero and the South Slavic Bugarštica: A Study of Repetitive Sequences and their Effect
on Narrative Style"). (Ph.D. dissertation, University of Chicago, 1973); idem. "Narrative Style
in Spanish and Slavic Traditional Narrative Poetry: Implications for the Study of the
Romance Epic," Olifant,2 (December 1974), pp. 109-128; idem, "The South Slavic Bugarštica
and the Spanish Romance: A New Approach to Typology," International Journal of Slavic
Linguistics and Poetics, 21 (1975), pp. 31-69; idem, "Medieval Spanish Epic and European
Narrative Traditions." La Corónica, 6 (Spring 1978), pp. 90-96; idem, "Oral-Traditional
Style and Learned Literature: A New Perspective," PTL: A Journal for Descriptive Poetics
and Theory of Literature, 3 (April 1978), pp. 345-356; idem, "Elaborate Style in South Slavic
Oral Narrative and in Kačić Miošić's Razgovor," in American Contributions to the Eighth
International Congress of Slavists, Zagreb andLjubljana, September 3-9, 1978, ed. Henrik
Birnbaum 2 vols. (Columbus, Ohio: Slavica, 1978), Vol. 1, pp. 522-531; idem, "Shamanistic
Features in Oral-Traditional Narrative," Language and Style, 11 (Fall 1978), pp. 223-225;
idem. "South Slavic and Hispanic Versified Narrative: A Progress Report on One Approach,"
in The Hispanic Ballad Today: History, Comparativism, Critical Bibliography, eds. Samuel
G. Armistead, Antonio Sánchez Romeralo, Diego Catalán, Romancero y poesía oral, no. 4
(Madrid: Cátedra Seminario Menéndez Pidal, Universidad Complutense de Madrid, 1979),
6 Olifant / Vol 9, Nos. 1&2/Fall & Winter 1981
tive style of the Roland differs considerably from that of oral-traditional or
folk poetry and indicate that the poem is most likely not an orally com-
posed text but a literary text which contains both written (or learned) as
well as oral or folk stylistic elements.
The various theories proposed thus far in order to account for the
mode of composition of the chanson de geste in general and the Roland in
particular, as the texts appear in extant manuscripts, offer a limited num-
ber of alternatives. These, briefly stated, may be reduced to three positions,
which are suggested by the long-standing traditionalist-individualist
debate and Lord's work on oral narrative tradition:8 (1) the proponents of
the traditionalist theory consider the poems to be direct products of an oral
tradition, the latter existing either according to Ramón Menéndez Pidal's
understanding of a memorized text circulating in variants or to the notion
of oral composition developed by Parry and Lord; such texts may have
been written down by a scribe under dictation, the "oral-dictated text," or
by the poet himself, the "autograph oral text";9 (2) according to the indi-
pp. 131-135: idem. "Hispanic and South Slavic Traditional Narrative Poetry and Related
Forms: A Survey of Comparative Studies (1824-1977)," in Oral Traditional Literature: A Fest-
schrift for Albert Bates Lord, ed. John Miles Foley (Columbus, Ohio: Slavica, 1981), pp.
375-389: idem, "Oral Literature and 'Pučka Književnost'; Toward a Generic Description of
Medieval Spanish and Other Narrative Traditions," in Folklore and Oral Communication,
ed. Maja Bošković-Stulli (Zagreb: Zavod za istraživanje folklora, 1981), pp. 155-166; idem,
review of Heroic Epic and Saga: An Introduction to the World's Great Folk Epics, ed. Felix J.
Oinas. Olifant, 7 (Spring 1980), pp. 300-302; idem, "Repetition and aesthetic function in the
Poema de mio Cid and South-Slavic oral and literary epic," Bulletin of Hispanic Studies, 58
(July 1981), pp. 189-196. Other briefer items in which Mitetich discusses his work appear as
follows: "Dissertation Abstract," Olifant,2 (December 1974), pp. 146-147 and Dissertation
Abstracts International, Vol. 36, No. 12 (1975-76), p. 8104-A; "1974 Annual Meeting of the
Société Rencesvals, American-Canadian Branch: Proceedings," Olifant,2 (February 1975),
pp. 164-166, 172-173; "Société Rencesvals Discussion Session on History. Narrative, and Dic-
tion in the Late Castilian Epic: Trends in Contemporary Research, December 28, 1977." Oli-
fant, 5 (March 1978), pp. 252-253, 262-265.
8For accounts of the traditionalist-individualist debate, see Martin de Riquer, Les
Chansons de geste françaises, French trans. Irénée-Marcel Cluzel, 2nd ed. (Paris: Nizet, 1957),
pp. 34-52; Ramón Menéndez Pidal, La Chanson de Roland et la tradition épique des Francs,
French trans. Irénée-Marcel Cluzel, 2nd ed. (Paris: Picard. 1960), pp. 3-82; W. G. van Emden,
'"La bataille est aduree endementres': Traditionalism and Individualism in Chanson-de-
geste Studies," Nottingham Mediaeval Studies, 13 (1969), pp. 3-26; C. W. Aspland, A Syntac-
tical Study of Epic Formulas and Formulaic Expressions Containing the -ant Forms in
Twelfth Century French Verse (St. Lucia: University of Queensland Press, 1970), pp. 1-38;
Duggan. pp. 1-6; Charles B. Faulhaber, "Neo-traditionalism, Formulism, Individualism, and
Recent Studies on the Spanish Epic," Romance Philology, 30 (August 1976), pp. 83-101. For
Lord's views, see Lord, The Singer, pp. 124-138.
Carton /Oral-Traditional Style in the Roland 7
vidualist view, the chansons de geste were composed in writing by learned
poets who made use of legendary material merely as a source of inspira-
tion;10 (3) finally, another group of scholars favors a compromise between
the two foregoing theories; according to them, the poems may be composi-
tions which were written in the style of an oral-traditional poetry;11 for
Lord, such deliberate imitations are the work of writers trained in a liter-
To date, the method developed by Professor Miletich appears to be
the only one providing both the comparative data and workable defini-
tions lacking in oral-formulaic studies and required for a clear differentia-
tion between "oral-traditional" and "written" styles in terms of the three
positions described above.
Miletich bases his distinction between "oral style" and "written style"
on the relation between what he calls "elaborate style" and "essential
style." According to his definition, "elaborate style" consists of repetitions
involving a recurrence in idea which is unnecessary to the unfolding of the
narrative line and thus delays, to some extent, the flow of narrative or
descriptive information. These repetitions "contribute no fundamentally
new dimension themselves as might be the case were synonyms with dif-
ferent connotations used in their stead."13 Consequently, a repetition is
considered to delay the flow of narrative information when it can be
determined that the latter could have been relatively advanced by the addi-
tion of a significant nuance or idea under the same textual conditions or
constraints. An example is the repetition which occurs in the second hemi-
stich of verse 157 in the Roland: "Bels fut li vespres e li soleilz fut cler"
(The afternoon was beautiful and the sun was bright).14 In this verse, the
9Menéndez Pidal, pp. 51-82; Lord, The Singer, pp. 126-129.
10See, for example, Joseph Bédier, Les Légendes épiques: Recherches sur la formation
des chansons de geste,2nd ed., 4 vols. (Paris: Champion, 1914-21), especially Vol. 3, pp.
183-453, with regard to the Roland; idem. La Chanson de Roland commentée (Paris: Piazza,
1937), pp. 1-64; Maurice Delbouille, "Les Chansons de geste et le livre," in La Technique
littéraire des chansons de geste: Actes du Colloque de Liège (septembre 1957), Bibliothèque
de la Faculté de Philosophie et Lettres de l'Université de Liège, 150 (Paris: Société d'Édition
"Les Belles Lettres", 1959), pp. 295-428; Italo Siciliano, Les Chansons de geste et l'épopée:
Mythes, Histoire, Poèmes (Turin:Società Editrice Internazionale, 1968).
11For accounts of the "conciliatory" point of view between traditionalists and individu-
alists, see, for example, Duggan, pp. 3-4 and n. 7; Aspland. pp. 10-12, 17-19.
12Lord, The Singer, p. 133.
13Miletich, "Repetitive Sequences," p. 79.
8 Olifant / Vol 9, Nos. 1 & 2/Fall &Winter 1981
repetition of the copula, estre, is not necessary for the unfolding of the nar-
rative line, because, should the second instance be omitted, the narrative
information supplied would remain the same. Moreover, this omission
would also provide a metrical slot which could be filled with an intensifier
such as mult, thus adding a nuance to the narrative line: "Bels fut li
vespres e li soleilz mult cler" (The afternoon was beautiful and the sun was
bright). According to Miletich, the narrative mode in the new hypothetical
second hemistich would no longer be "elaborate", but "essential", i.e.,
characterized by narrative economy and necessity.15
This distinction between "elaborate style" and "essential style" stems
from a comparative study of the Spanish romance and the South Slavic
bugarštica, the purpose of which was to clarify the relationship between
narrative style and definition of genre. In his early work, Miletich stressed
that, from the standpoint of narrative style, the distinction between ballad
and epic, for instance, remained unclear in spite of a considerable amount
of research done on the subject. For example, Menéndez Pidal had estab-
lished that traditional poetry in general and the romance in particular
may be characterized on one hand by rapid style, i.e., a tendency to avoid
what is seen as non-essential information, and on the other by repetitive
diction, a feature he connected to the lyrical dimension of the romance,
which, according to him. distinguished the latter from epic proper, like
the chanson de geste. However, the Spanish scholar never accounted for
the exact nature of the relationship between these two seemingly opposite
tendencies.16 In an effort to provide a more accurate and systematic
approach to the definition of genre characteristics as regards the romance
and the bugarštica, Miletich undertook a study of types of closely occur-
ring repetitions appearing in a given text and of their effect in the narra-
tive style of the same text. He fixed the number of metrical units (e.g.,
hemistichs), within which a closely occurring repetition is to be found, at
seven, a limit chosen because it seemed to encompass most of the repeti-
tions in the texts under consideration and thus proved practical in gaug-
14 Raoul Mortier, ed., Les Textes de la Chanson de Roland, 10 vols. (Paris: Geste Fran-
cor, 1940-44), Vol. 1: La Versiond'Oxford.
15For discussions of the difference between "elaborate style" and "essential style," see,
for example, Miletich, "Narrative Style," pp. 115-116; idem. "Repetitive Sequences." p. 99;
Jean-Paul Carton, "Oral-Traditional Style and the Song of Roland: 'Elaborate Style' and
'Essential Style'" (Ph.D. dissertation, University of Utah, 1982), pp. 48-60.
16Rimón Menéndez Pidal, Romancero hispánico (hispano-portugués, americano y
sefardí): Teoría e historia,2 vols, (Madrid: Espasa Calpe, S.A.. 1953), Vol. 1, pp.59-80; cited in
Miletich, "The South Slavic Bugarštica," pp. 53-54, 67; idem, "Narrative Style," p. 116.
Carton / Oral-Traditional Style in the Roland 9
ing different repetitive tendencies in different kinds of texts and similar
tendencies in similar kinds of texts for the various meters studied so far.17
To these closely occurring repetitions, Miletich added the repetition of
extended groups of hemistichs ("repetitive groups") which may occur suc-
cessively anywhere in the same poem and are a significant element of nar-
rative style, because, to a greater or lesser degree, they retard the flow of
narrative or descriptive information. These groups are united by a com-
mon idea and thus correspond to some extent to Lord's "themes."18 The
classification of the repetitions thus observed by Miletich includes six
types, which may be divided into two groups, "elaborate style" repetitions
and "essential style" repetitions, according to whether or not they involve
the unnecessary recurrence of an idea. Briefly stated, the categories of
"elaborate style" repetitions in general terms are as follows: (1) repetitions
of groups of hemistichs which, as mentioned above, are not closely occur-
ring and correspond to some extent to Lord's "themes" (category Î:
"repetitive group"); (2) closely occurring repetitions of entire hemistichs,
lines or couplets in which the idea, diction, and syntax remain essentially
unchanged (category II: "exact repetition"); (3) closely occurring repeti-
tions in idea generally involving different diction and syntax (category III:
"semantic repetition"); and (4) closely occurring repetitions which involve
the same or similar diction but do not fill a metrical unit (category IV:
"similar repetition"). "Essential style" repetitions are: (1) closely occurring
repetitions which involve the same or similar diction but no unnecessary
recurrence in idea (category V: "distinct repetition"); and (2) closely occur-
ring repetitions of grammatical patterns (category VI: "syntactic repeti-
tion").19 Miletich's preliminary studies of the foregoing repetitions,
which he refers to as "repetitive sequences," showed that the rapid or
"essential" style of the romance described by Menéndez Pidal was indeed
accompanied, like that of the bugarštica, by a measurable opposite effect,
i.e., "elaborate style," which appeared to be another important aspect of
Through further extensive analyses of the effect of repetitions in the
style of both authentically oral poems and written imitations of oral
17See, for example, Miletich, "Narrative Style," p. 111.
18Miletich, "Narrative Style," pp. 111-113.
19Ibid., pp. 112-115. For examples of repetitions belonging to categories II through V
in the Roland, see below, the notes on pp. 14-17 and corresponding material in text.
20Miletich, "Narrative Style," p. 116.
10 Olifant / Vol 9, Nos. 1&2/Fall & Winter 1981
poetry, Miletich was also able to observe that the former displayed a higher
instance of "elaborate" repetitions and, thus, he devised a method which
could give a quantitative account of such tendencies. A ratio of the
"elaborate style" to the "essential style" in a given narrative poem can be
provided by calculating the number of metrical units which involve
"elaborate style" repetitions and measuring it against those which contain
"essential style" repetitions in addition to non-repetitive units, the
"remainder," which together constitute "essential style."21
first and then of longer works have shown a consistent pattern. "Elaborate
style" appears to be a distinctive feature of the traditional or folkloric
material, either memorized and circulating in variants according to
Menéndez Pidal's theory, or orally composed according to Lord's concept.
The shorter texts examined in this category are composed both of authen-
tically oral songs and of narrative poems of a more debatable authenticity
which have been considered oral by a good number of scholars. Among the
former are modern Judeo-Spanish romances of Morocco, modern Serbo-
Croatian heroic songs, the earlier Serbo-Croatian songs from the classic
Vuk Karadžić collection (volume II) and nineteenth-century Russian by-
liny.The latter include sixteenth-century romances of the Wolf and Hof-
mann collection and eighteenth-century bugarstice. In addition to this
material, Miletich has also investigated "elaborate style" in a larger narra-
tive poem, the orally composed epic, the Song of Bagdad, of which Lord
has analyzed a fifteen-line sample.22 In all of these texts, "elaborate style"
repetitions represent roughly an average of one-third of all metrical units
considered and tend to be evenly distributed, the Song of Bagdad being
slightly more "elaborate" than the average for all of the oral-traditional
material studied thus far (37 percent). "Elaborate style" thus appears to
constitute an essential element of the weave of oral-traditional narrative
poetry and, indeed, analyses of poems known to have been composed in
writing show that, although "elaborate" repetitions may be present in
such texts, they are far from being so pervasive an element of style as in the
case of the material discussed above. For example, the eighteenth-century
poems of Andrija Kačić Miošić, a Franciscan monk, who, according to
Lord, knew the oral epic of the South Slavs very well but composed his
works in writing, show an average of 13 percent "elaborate style," the rep-
21 Miletich. "Medieval Spanish Epic." p. 91.
22Lord, The Singer, p. 46.
Carton / Oral-Traditional Style in the Roland 11
etitions tending to be clustered in certain parts of the poems.23 An analysis
of a "written" poem by Lorca displays 5.4 percent "elaborate style."24
In an application of the method to medieval Spanish epic poetry,
Miletich found that the entire Cantar de Mio Cid and the entire Mocedades
de Rodrigo yield only 17 and 19 percent "elaborate style," respectively, and
that the repetitions do not tend to be evenly distributed throughout the
texts. The Cid and the Mocedades thus appear to be closer to Kačić
Miošić's work, both in the percentages and the distribution of their
"elaborate style" repetitions. On the basis of the foregoing evidence, Mile-
tich concludes that these poems correspond to a genre which Slavic schol-
ars refer to as pučka književnost, i.e., a kind of literature using elements
belonging to both oral tradition and learned tradition. This category,
which displays an average of about 16 percent "elaborate style" and a spo-
radic distribution of the repetitions,25 therefore pertains to the texts
described in section (3) of the classification presented in the second para-
graph of the present article. To be noted here is that Miletich's conclusion
is corroborated by Alan D. Deyermond's generally accepted (at least by
American hispano-medievalists) theory regarding the learned authorship
of the Mocedades.26
Although more work is being done by Miletich on quantitative
analysis of "elaborate style" in narrative poems, on the basis of the forego-
ing results, his method already appears to provide an adequate tool for
testing the orality or the literacy of ancient and medieval narrative poetry.
In contrast to formulaic studies in general and Duggan's work in particu-
lar, Miletich's differentiation between oral style and written style in narra-
tive poetry rests on extensive evidence and a posteriori reasoning. As
already stated, formulaic studies to date have not produced extensive data
which may enable scholars of narrative poetry to distinguish clearly
between oral and written styles.27 This is perhaps the single main flaw of
Parry-Lord studies and their application to medieval epic poetry as a tool
23Miletich, "Medieval Spanish Epic," pp. 91-93.
24Miletich, "Stilističke razlike," p. 126; I acknowledge the assistance of John S. Miletich
for the information in Serbo-Croatian.
25Miletich. "Medieval Spanish Epic," pp. 92-93.
26Alan [D.] Deyermond, "The Mocedades de Rodrigo as a Test Case: Problems of
Methodology," La Corónica, 6 (Spring 1978), pp. 108, 111 and n. 2.
27See above, note 5.
12 Olifant / Vol. 9, Nos. 1&2/Fall &Winter 1981
for the investigation of the orality or literacy of extant texts. Miletich, on
the other hand, developed his criteria from a statistical and descriptive
examination of some 15,000 verse lines of authentically oral and written
texts as well as written imitations of oral style in two basically different
language families, the Romance and the Slavic, comprising Spanish,
Serbo-Croatian, and Russian-language material.28 Furthermore, these
data have been obtained in an application of the method to both oral and
written texts in different literary and linguistic traditions, so that the
assessment of the mode of composition of a given problematic text does
not rest solely on analogy but also on observations made in each of the
main language traditions examined by Miletich, namely Spanish and
Serbo-Croatian.29 Therefore, although the Old French epic cannot be
compared directly to an authentic French oral heroic tradition because of
the lack of such primary material in French, the application of this
method to poems like the Chanson de Roland is validated by the fact that
Miletich's studies show that the observed phenomena (density and distri-
bution of "elaborate style") do not depend on a particular linguistic group
but appear consistent in the investigation of the two basically distinct lan-
guage families studied so far. Finally, another point of importance which
must also be mentioned here is that, also in contrast to the Parry-Lord
theory, the distinction between oral style and written style does not rest on
an assumption, namely that a higher percentage of a particular kind of
repetition is an intrinsic element of oral style.30 Although Miletich has
addressed the problem of the origin of the repetitions he studies,31 the sta-
tistical method itself and the classification criteria as they stand to date are
not concerned with or based upon an examination of why the categories of
"elaborate style" repetitions occur in greater quantity and more evenly in
an oral narrative poem than they do in a written poem but merely point
out, on the basis of extensive observations, that such is the case.
From the standpoint of the percentage and distribution of "elaborate
28Miletich, "Repetition and aesthetic function," p. 189.
29Miletich, "Medieval Spanish Epic," p. 93.
30For questions raised concerning the relationship between the "formula" and oral
composition, see Delbouille, pp. 343, 362; Larry D. Benson, "The Literary Character of
Anglo-Saxon Formulaic Poetry," PMLA [Publications of the Modern Language Associa-
tion], 81 (October 1966), pp. 334-341.
31 Miletich, "Medieval Spanish Epic," p. 92 and n. 11; idem, "Shamanistic Features,"
Carton / Oral-Traditional Style in the Roland 13
style" repetitions, the narrative style of the Raoul Mortier edition of the
Oxford version of the Chanson de Roland32resembles that of comparable
texts classified by Miletich as pučka književnost. Here, I would like to
emphasize that great care was devoted to making the figures proposed in
the present study directly comparable to those obtained by Miletich for the
Spanish, Serbo-Croatian, and Russian texts he investigated. No adjust-
ment which might constitute a breach of the method was made, and much
of the data was reconsidered in the light of Miletich's recommendations in
order to ensure that the application be consistent in all respects.
That the Oxford manuscript is employed for such a study is only fit-
ting because Roland scholarship has consistently turned to this version,
the oldest Old French epic text, for problems dealing with the origins and
the nature of the poem in particular and the chansons de geste in general.
The choice of the Mortier edition is determined by Duggan's use of that
text. It not only made Duggan's very helpful concordance of the poem
more immediately accessible to me,33 but will also validate any future
comparison with his results. This factor is of some importance because the
principal problem addressed in the present study has its point of departure
in Duggan's work. As for Mortier's "errors," I also used Cesare Segre's crit-
ical edition and John Robin Allen's unpublished corrections of the Mor-
tier text in order to ensure that they would not significantly alter the
results of my study. In fact, these corrections do not affect the analysis of
"repetitive sequences" in the text used by Duggan.34
The Roland may be compared directly with the longer texts studied
by Miletich, namely the Cantar de Mio Cid, the Mocedades de Rodrigo and
the authentically oral epic, the Song of Bagdad. I have determined its
"elaborate style" content to be 20.48 percent, a figure significantly lower
than the one arrived at by Miletich for the oral poem (37 percent) but close
to the "elaborate style" content of the two medieval Spanish texts, the Cid
(17 percent) and the Mocedades (19 percent), which, as mentioned earlier,
are classified as pučka književnost. Moreover, also like the Cid and the
Mocedades, but unlike the oral-traditional poems analyzed by Miletich
32See above, note 14.
33Joseph J. Duggan, A Concordance of the "Chanson de Roland" (Columbus, Ohio:
Ohio State University Press, 1969).
34Cesare Segre, ed., La Chanson de Roland (Milano, Napoli: Riccardo Ricciardi, 1971);
John R. Allen, "On the Mortier Edition of the Chanson de Roland." I am indebted to Pro-
fessor Allen for providing me with a copy of this manuscript.
14 Olifant / Vol. 9, Nos. 1&2/Fall & Winter 1981
(both the longer and the shorter ones), the Roland displays an irregular
distribution of its "elaborate style" repetitions.35 The latter are clustered
in certain laisses or scenes and are absent or almost absent from others. In
order to illustrate this sporadic character of the distribution of the repeti-
tions in the Old French poem, I have selected a passage which clearly
shows different tendencies, laisses LIX through LXVIII, i.e., 109 verses
which begin with Roland's reaction to his nomination as the leader of the
rear guard by Ganelon. In this passage, I have marked all instances of
"elaborate style" repetitions, placing the first occurrence of repetitive ele-
ments into parentheses and putting the recurrence(s) into italics.
"Essential style" repetitions are left unmarked.
Li quens Rollant, quant il s'oït (juger), AOI.
Dunc ad parled a lei de chevaler:
«Sire parastre, mult vos dei aveir cher:
La rereguarde avez sur mei jugiet.36
755 (N'i perdrat Carles), li reis ki France tient,
Men escientre palefreid ne destrer,
Ne mul ne mule que deiet chevalcher,
Ne n'i perdrat37ne runcin ne sumer,
Que as espees ne seit einz eslegiet.»
760 Guenes respunt: «Veir dites, jol sai bien.» AOI.
Quant ot Rollant qu'il en en la rereguarde,
Ireement parlat a sun parastre:
«Ahi! culvert, malvais hom de put aire.
Quidás le guant me caïst en la place,
765 Cume fist a tei le bastun devant Carle?» AOI.
35For the complete data, see Carton, pp. 166-187; ibid., pp. 151-156 for a statistical
36Category IV: "similar repetition" (recurrence of identical or similar words shorter
than the metrical unit).
37Category II: "exact repetition" (recurrence of at least one entire metrical unit). To be
noted is that "exact repetition" does not necessarily mean verbatim repetition. Variation may
occur as long as the idea contained in the units matched remains basically the same and is
expressed through essentially the same diction. Such variation often involves, for example,
less semantically weighted elements and morphological changes but may sometimes also
consist of the omission of a main word as is the case in the "repetitive sequence" of vv.
755-758: the word Carles, which appears in the first hemistich of v. 755, does not recur in the
first hemistich of v. 758 and the syllable left vacant by this omission is filled by a less semanti-
cally weighted element, the coordinating conjunction ne, in such a way that both hemistichs
in their entirety express the same idea, i.e., "Charles will not lose." For a discussion of varia-
tion in category II, see Carton, pp. 115-120.
Carton / Oral-Traditional Style in the Roland 15
— «Dreiz emperere.» dist Rollant le barun,
«Dunez mei l'arc que vos tenez el poign.
Men escientre nel me reproverunt
Que il me chedet, cum fist a Guenelun
De sa main desire, quant reçut le bastun.»
Li empereres en tint sun chef enbrunc,
Si duist sa barbe e detoerst sun gernun;
Ne poet muer que des oilz ne plurt.
Anpres iço i est Neimes venud:
Meillor vassal n'out en la curt de lui;
E dist al rei: «Ben l'avez entendut;
(Li quens Rollant), il38 est mult irascut.
La rereguarde est jugée sur lui:
N'avez baron ki jamais la remut.
(Dunez li) l'arc que vos avez tendut,
Si li truvez ki tres bien li aiut!»
Li reis li dunet, e (Rollant) l'a reçut.
Li empereres apelet (ses nies) Rollant:
«Bel sire nies, or savez veirement.
Demi mun host vos lerrai en present.
Retenez les, ço est vostre salvement.»
Ço dit li quens: («Jo n'en ferai nient;
De us me cunfunde, se la geste en desment!
.XX. milie Francs retendrai ben vaillanz.
Passez les porz trestut soürement:
Ja mar crendrez nul hume a mun vivant!»)
(Li quens) Rollant est muntet el destrer. AOI.
Cuntre lui (vient) sis cumpainz Oliver;
Vint (i) Gerins e li proz quens Gerers,
(E) vint i Otes, si i vint Berengers
E vint i Astors e Anseïs li veillz;
Vint i Gerart de Rossillon li fiers;
Venuz i est li riches dux Gaifiers.
(Dist) l'arcevesque: «(Jo) irai par mun chef!»
— («E jo od vos,») ço dist (li quens) (Gualters);
38Category III: "semantic repetition" (recurrence of an idea through different diction).
Here, the repetition consists of the pleonastic use of the subject pronoun il, which repeats Li
quens Rollant. Other similar instances of pronouns used pleonastically occur in v. 787.
where ço anticipates Roland's words (vv. 787 [second hemistich]-791), and vv. 800 (ço)and
16 Olifant / Vol.9, Nos. 1&2/ Fall &Winter 1981
«Hom sui (Rollant), jo39 ne li dei faillir.»
Entr'els eslisent .XX. milie chevalers. AOI.
Li quens Rollant Gualter40 de l'Húm apelet:
«Pernez mil (Francs) de France,41nostre tere,
Si purpernez les deserz e les tertres,
Que l'emperere nis un des soens n'i perdet.» AOI..
Respunt (Gualter): «Pur vos le dei ben faire.»
Od mil (Franceis) de France, la lur tere,
Gualter desrenget les destreiz e les tertres:
N'en descendrat pur malvaises nuveles,
Enceis qu'en seient .VII. C. espees traites.
Reis Almaris, del regne de Belferne,
Une bataille lur livrat (le jur) pesme.
Halt sunt li pui e li val tenebrus,
Les roches bises, les destreiz merveillus.
Le jur passerent Franceis a grant dulur;
De .XV. lius en ot hom la rimur.
Puis que il venent a la Tere Majur,
Virent Guascuigne, la tere lur seignur,
Dune le remembret des fius e des honurs,
39Category IV: according to Miletich, the repetition of less semantically weighted ele-
ments such as function words may also be "elaborate" if it can be determined that an adequate
new idea or nuance could have been used in its place. In v. 801, the pronoun jo could have
been replaced by a short adverb such as ja or or.
40V. 803 contains three repetitions of words which occur first in the second hemistich of
v. 800 and the first hemistich of v. 801, li quens, Rollant and Gualter. The repetitions of li
quensand Rollant are "elaborate" (category IV) because the hemistich in which they appear
may be replaced in its entirely by Li nies Carlun (linies replacing li quens and Carlun replac-
ing Rollant),which could thus present Roland in a different light. On the other hand, in the
second hemistich of v. 803, the repetition of Gautier's name is "essential" (category V:
"distinct repetition") because the language of the text does not provide an expression which
could have fulfilled the same function. This repetition is necessary for contextual identifica-
41Category IV: it must be noted, however, that my decision to classify this repetition as
"elaborate" is contrary to the opinion of Lucien Foulet, who sees in the expression Franc de
France a possible opposition between the general sense of Franc, meaning someone from
anywhere in the whole empire and the restrictive sense of France, meaning France proper: "il
faut bien que l'expression Francs de France exclue les Francs qui ne sont pas de France"
(Lucien Foulet, Glossary of the Oxford Version of the Chanson de Roland, in Bédier, La
Chanson de Roland commentée, p. 511, col. 2). However, Foulet does not give any indication
as to the repetition of v. 1593, D'Affrike i ad un African venut, or that of v. 3038, Alemans
sunt e si sunt d'Alemaigne, both of which are more clearly pleonastic. Because no definite
answer can be provided for the repetitions of vv. 804 and 808, I classified them as "elaborate"
as a precaution against an underestimation of "elaborate style."
Carton / Oral-Traditional Style in the Roland 17
E des pulcele e des gentilz oixurs:
(Cel nen i ad ki de pitet ne plurt.
Sur tuz les altres est Carles anguissus:)
As porz (d'Espaigne) ad lesset sun nevold,
Pitet l'enprent, ne poet muer n'en plurt.42AOI.
Li .XII. per sunt remes en Espaigne.
.XX. milie Francs unt en lur cumpaigne,
N'en unt poür ne de murir dutance.
Li emperere s'en repairet en France;
Suz sun mantel en fait la cuntenance.
Dejuste lui li dux Neimes chevalchet
E dit al rei: «De quei avez pesance?»
Carles respunt: «Tort fait kil me demandet!
Si grant doel ai ne puis muer nel pleigne.
Par Guenelun serat destruite France:
Enoit m'avint un avisiun d'angele,
Que entre mes puinz me depeçout ma hanste,
Chi ad juget mis nes a rereguarde.
(Jo) l'ai lesset en une estrange marche!
Carles li magnes ne poet muer n'en plurt.
.C. milie Francs pur lui unt (grant) tendrur,
E de Rollant merveilluse poür,
Guenes li fels en ad fait traïsun:
(Del rei paien) en ad oüd granz duns,
Or e argent, palies e ciclatuns,
Muls e chevals e cameilz e leuns.
Marsilies mandet d'Espaigne les baruns,
Cuntes, vezcuntes e dux e almaçurs,
Les amirafles e les filz as cunturs:
.IIII. C. milie en ajustet en .III. jurz.
En Sarraguce fait suner ses taburs;
Mahumet levent en la plus halte tur.
N'i ad paien nel prit e nel aort.
Puis si chevalchent, par mult grant cuntençun,
La Tere Certeine e les vals e les munz:
De cels de France virent les gunfanuns.
La rereguarde des .XII. cumpaignuns
Ne lesserat bataille ne lur dunt.
The "elaborate style" content of this passage is approximately 11.93 per-
42Category III: "semantic repetition" (v. 825 is "elaborate" because of vv. 822-823, which
already show Charlemagne as one among those who are weeping).
18 Olifant / Vol 9, Nos. 1 & 2/Fall &Winter 1981
cent (26 "elaborate" hemistichs / 218 hemistichs), a figure which is signifi-
cantly lower than the average for the entire poem (20.48 percent) and
slightly below the figure of 13 percent found by Miletich in Kačić Miošić's
written compositions (shorter poems classified under pučka književ-
nost).43As emphasized by the italics, "elaborate style" repetitions appear
quite unevenly. Half of the 26 repetitive units are clustered in two laisses,
laisses LXIV (9 hemistichs) and LXV (4 hemistichs). Two of the remain-
ing laisses, laisses LX and LXI, contain no "elaborate style" repetitions,
and elsewhere the repetitions are either isolated (e.g., v. 840) or grouped in
small clusters (e.g., vv. 782-784). This ten-laisse excerpt is not unique, but
rather characteristic of the style of the Roland. For example, "repetitive
groups" (category I), the category of extended repetitions which are not
closely occurring and correspond more or less to Lord's "themes," are
especially concentrated in laisses parallèles such as those describing boasts
or single combat scenes, as well as in laisses similaires. Moreover,
although the average percentage of "elaborate style" may in certain pas-
sages resemble figures obtained in the analysis of oral-traditional material,
elsewhere, in other sections of comparable length, it is in fact quite low.
For example, verses 1188 to 1296, a passage of 109 verses involving a series
of laisses parallèles and taken from the first encounter between Charle-
magne's rear guard and the Saracens, contain approximately 41.76 percent
"elaborate style," including 26 percent category-I repetitions. These fig-
ures are comparable to the 37 percent "elaborate style" and the relatively
high percentage of category-I repetition found by Miletich in the analysis
of the oral Yugoslav epic, the Song of Bagdad.44On the other hand, the
last 111 verses of the Roland (vv. 3892-4002) display only about 6.31 per-
cent "elaborate style" and do not include a single "repetitive group."45
Although my analysis of "elaborate style" in the Chanson de Roland
argues that the Oxford version of the poem is not an oral-traditional text
but rather, in all probability, a learned text composed in writing, I must
stress that this conclusion, which supports the position adopted by schol-
ars who favor a compromise between individualist and traditionalist theo-
ries, does not inform us about the origins of the poem, but rather solely
about its composition. The results I have obtained suggest a pre-existing
oral tradition, such as the one the historical traditionalists have tried to
43See above, note 23.
44Miletich, "Medieval Spanish Epic," p. 92.
45For "elaborate style" in vv. 1188-1296 and 3892-4002, see Carton, p. 162, n. 14.
Carton / Oral-Traditional Style in the Roland 19
reconstruct, but reveal little of its nature and do not even indicate that the
learned poet had direct contact with it. They merely point out the extent to
which Turoldus (?) has made use of certain oral-traditional stylistic
devices which, insofar as we know, may have been transmitted to him
either orally or in writing.46