I have been especially interested in the fact that there are so many variants of the dainas. With these poems two questions arose: If these are ritual songs—and many of them are—they should, as we are sometimes told, be word-for-word exact in order to satisfy the ritual requirements. They should be like incantations, which are supposed not to work unless repeated exactly. Then how come the variants? The second question: If it is true that short forms are easily memorized, one would expect that eight lines would be stable, so how come the variants? Something else seems to be at work here, and one has to ask the questions again: Is it necessary to postulate word-for-word exactness, or is that making an untrue assumption with reference not to reality but only to a guesswork theory based on a misunderstanding of what the practitioners of the form have said under inept questioning? Is shortness a real criterion for memorization, or is another process of transmission at work here?

The dainas are short and for the most part nonnarrative. The quatrain is their favored form. An excellent example of a daina is the following, Ltd 1148:

The parallelisms between the two distichs are typical of daina structure. There is a variant of this daina, Ltd 1148.1:

The parallelisms of the distichs tie the quatrain closely together.

As an illustration of a six-line daina I have chosen one with many variants. It is usually included in the mythological cycle.

Nos. 33855-59 begin with essentially the same quatrain.

The variations appear in the last couplet of the six-line stanza (or in the last quatrain of No. 33859), a fact that may explain the star’s absence. Barons considered these four dainas as type-songs, but we would class them as variants of one another, or simply variants.

The five dainas just cited demonstrate the importance of sound patterns in the Latvian oral traditional songs, a trait they share with all other traditional poetry. Note for example the persistent v alliteration, as in lines two and three, “vai ir visas vakarā. / Ira visas vakarā,” where it is also coupled with i and a assonance. The v alliteration had begun actually in the first line in zvaigznes ‘stars’, the key word for the daina, as does the assonance; and it continues in line four in vien nevaid, and in 33856, 33858, and 33859 it appears again in Vāczemē, perhaps even having some influence in the choice of that word. In 33859 it is carried along for two additional lines, “Saulītei svārkus šuva [where it combines with the “sun” word saulītei] / Vienu strīpi zelta lika.”

The second line of the distich 27307-17, Es vairs ilgi nedzīvošu, “I shall not live long any more,” is used in 27307-12, and with Nu es (now I) substituted for Es vairs in 27313-16, and with Nu vairs (now) in 27317. It also has a variant, Ka es (that I), which is used in either the second or the fourth line of a quatrain:

and

The dainas are oral traditional formulaic compositions that exist in many variants. They have their own traditional diction and style, and when someone composes a daina not in that style, the resulting song is not felt to be “right.”

If one looks at a type-song and its variants, one is struck by the fact that they frequently share a more or less stable core of lines, or even of couplets, which themselves can be varied within given limits and to which other lines and couplets are added. Although Barons recorded the places in which his {30|31} songs were collected and published a list of people who sent them to him, we do not know from whom the individual songs were collected; it is patent that they are not from one individual singer. The variations, therefore, are not in the singing of the same song by the same singer. It is correct to say that the song as we see it in available variants has a more or less stable core, provided that we realize that there is no implication that this statement is necessarily true for the song as it exists in the practice of any one singer.

Collectors in this century are meticulous in recording information about the informants, the storytellers and singers, from whose creative minds they have taken down stories and songs. This careful concern for method makes it possible for scholars to investigate the workings of the individual artist as well as of the tradition as a whole. The Milman Parry Collection, like other modern collections, set down such information, and it is to it, and to other collections like it in its rigorous methodology, that we must turn for evidence of the practice of individual singers in varying or not varying the songs they have heard.

It is usually assumed that, because they are short, lyric songs are memorized verbatim. But as with the Latvian dainas, the number of variants would seem to indicate that the texts are not as fixed as memorization would suggest. I pay particular attention here to the concept of textuality as a means of determining whether the South Slavic songs are fixed textually and hence {33|34} memorized, with or without writing, or whether their texts are fluid and the result of composition in performance. I note the degree of verbal correspondence among the variants of the lyric songs, as I have earlier, in The Singer of Tales, among the variants of a “theme,” or repeated passage, in the investigation of the composition and transmission of the South Slavic epic songs. For we seek to discover whether or not there is a principle of composition and transmission which is to some extent shared by both oral traditional epic and oral traditional lyric songs in the South Slavic poetries.

In some of the songs and variants the first part of the song is comparatively stable and the variants diverge in their endings. Take, for example, Mladenović 155:

There are two variants of this song in Karadžić Vol. 1. The first is 1.590:

I have underlined the differences in the first six lines between Nos. 590 and 155. After line six they diverge considerably.

The second variant is 1.591: {34|35}

The underlinings are of the differences in the first four (or eight, respectively) lines between 591 and 155. The songs diverge greatly after line four (or eight). In the first part of the song the differences between the variants are negligible—ljubičice ‘violet’ in 591 instead of ružičice ‘rose’ in 155 and 590; besediš (besjediš) and besedila (besjedila) in 590 and 591 instead of govoriš and govorila. Song 155 and the first half of song 591 end abruptly with a one-line curse. In 590 and 591 the singer develops the theme of “mother” (590) or (591) elaborates on who the “they” are who are either forbidding or allowing her to speak, as mentioned in 155. There is, then, a stable part of the song which may be continued in different ways. In this case the stable part of the song is the beginning. A sense of textuality belongs to that part but not to the rest of the song.

In the second part of the song, Mladenović 155 and Karadžić 1.591 add to the opening six lines, which end in each case with al’ mi ne daju ‘but they do not let me,’ ko to ne da, ko l’ to smeta ‘who does not let you,’ ‘who interferes,’ and they diverge only in the final five-syllable coda: Bog nam ga smeo! ‘May God take him away!’ in 591 and Ubioga Bog! ‘May God strike him down!” in 155. That is where Mladenović 155 ends; undoubtedly Karadžić did not publish it for this reason, considering the other two fuller and, hence, better.

Karadžić 591 answers the question who will not allow the girl to speak. Vuk Karadžić 590 does not ask the question at all, because the girl states ne smem od majke ‘I dare not because of my mother’. The two endings have nothing in common. There is a sense of textuality in the openings, but the second parts are different songs, and one would have to look elsewhere, in some other songs, for their counterparts. One can speak here of the mixing of songs. {35|36}

In our next example the stable part of the song comprises its middle and final sections, whereas the beginnings exhibit considerable variety, as was the case with the endings in the previous example. Mladenović 143 has four variants—Mladenović 144 and Karadžić 1.285, 1.286, and 5.379. The stable parts consist of a series of questions and the answers to them. The settings of the questions and answers vary from song to song. Here are the questions and answers in Mladenović 143 and 144.

One of the differences between these two texts is that Mladenović 143 is in octosyllables (3-2-3) and 144 is in decasyllables (4-6). Of the five questions asked in 143 only four are answered. ‘Wider’ (sire), ‘deeper’ (dublje), and ‘swifter’ (brže), in that order, form three of the five questions in both. ‘Dearer’ (milije) in 143 is represented by draže in 144, and it characteristically ends the series. The fourth question is different in each—’brighter’ (sv’jetlje) in 143 and ‘sweeter’ (slađe) in 144. Of the objects in the questions, only ‘field’ (polje) and ‘sea’ (more) are found in both songs, but in reverse order. The object in the fifth question is always a member of the family—’brother’ (brat) in 143 and ‘mother’ (mater) in 144. ‘Swifter than a horse’ (brže od konja) in 143 is matched by ‘swifter than a gray falcon’ (brže od siva sokola) in 144. ‘Brighter than a sword’ (sv’jetlje od mača) in 143 and ‘sweeter than rose conserve’ (slađe od đulbe šećera) in 144 have no counterparts in the other song. Some of these differences result from the difference of meters.

The answers vary more than the questions. In the following translations I have underlined the same or similar objects that are wider, deeper, swifter, brighter, sweeter, or dearer than another object.

The textuality in the questions and answers sections of these two songs is not as taut as that in the beginning portions of our previous examples. Greater variety is allowed. But the textuality is real nevertheless.

Let us look at the other three variants. First Karadžić 5.379 (lines 19-29):

I have underlined the elements found in the previous two variants. The order is different. The horse has an appositive/epithet. There is metathesis of noun and epithet in mora sinjega of this version with the more usual sinjega mora of Mladenović 144. The family members of the last question are “father and mother” rather than “brother” or “mother.” Only one line in the question-and-answer series is different from the other two, and in the question part, ‘better’ (bolje) is inappropriate—a mistake, in fact—as the answer, ‘whiter’ (bjelji) shows. In spite of these differences, the sense of textuality, of certain specific words, is strong in all three variants.

The last two variants in Karadžić are:

No. 285, it is to be noted, is octosyllabic and has five questions and answers, whereas 286 is decasyllabic and has only three. In the other texts of this song in Karadžić, there is only one line between the two quotations, but 286 is an exception with four lines. The comparatives and the objects in 286 are, nonetheless, to be found in the other texts, each one of which, however, has some unique element. In 285, that element is “honey.”

There can be no doubt about a sense of textuality in these sections of the five variants, although it must be stressed that that does not argue for a fixed text and memorization. What it does exhibit is a remembering of a number of “more or less fixed” texts and a selection, conscious or unconscious—probably the latter—of elements from them.

The variables in the “more or less stable” text in the case of this “theme” (repeated passage) can be stated in the following way. The comparatives ‘wider’ (šire), ‘swifter’ (brže), and ‘dearer’ (either draže or milije) are the most likely to appear. ‘Sweeter’ (slađe), ‘longer’ (dulje), ‘higher’ (više), and ‘deeper’ (dublje) may also be found. ‘Brighter’ (sv’jetlje) and ‘whiter’ (bjelje) are used once each in our five texts, a fact that indicates that other less common comparatives are possible. Not only is there a choice of the usual comparatives as they may be known to a given singer, but there is also an opportunity for departing from the trodden path. The only obligatory elements are {39|40} ‘dearer’ (draže) and one other comparative to introduce it. These comparatives can be used whether the meter is octosyllabic or decasyllabic.

The nouns that follow the comparative are somewhat limited by meter as well as by traditional usage. For example, one cannot say ‘swifter than a falcon’ (brže od sokola) in the octosyllabic line (3-2-3) because od sokola has one syllable more than is allowed. Konj ‘horse’ is useful here, because od konja fits perfectly into the last position in the 3-2-3 line, whether it be “Što li je brže od konja” or “Ima ľ što brže od konja.” Majka ‘mother’ is possible in an octosyllabic line (although it is not used in our texts), but mater ‘mother’ is not possible, because its genitive, matere, produces an extra syllable. In our two octosyllabic songs, brat ‘brother’ is used for the family member. Med ‘honey’ does not, to the best of my knowledge, have a traditional epithet and cannot be used in the appropriate position, that is, in the second part, of a decasyllable, but it fits perfectly into the octosyllable. There is a rare noun going with ‘brighter’ (sv’jetlje) in the octosyllable, namely, ‘sword’ (mač). These are the “more or less stable” components in the octosyllable.

The variables in the decasyllable overlap with those in the octosyllable, because those found in the octosyllable can be adapted for the decasyllable by the use of an epithet. Thus we have the following:

There are four nouns in this position which are confined to the decasyllable.

These are some of the variables in this position in the decasyllable that define a “more or less stable” text.

Finally, what are the variables for the other noun, the object or person better, bigger, whiter, or swifter, for example, than something or someone else? What is needed is a two-syllable noun in the first part of the decasyllable (4-6) or the middle two syllables in the 3-2-3 octosyllable. The same noun {40|41} will fit either meter. The most common are more ‘sea’, nebo ‘sky’, oči ‘eyes’, and dragi ‘beloved’. The rare ones are:

The texts that we have seem to be the result of remembering known and used variables rather than of memorization of a fixed text.

If the questions and answers display textuality, that cannot be said for the beginnings of the songs, which describe the characters and the settings of the questioning. I shall review these before commenting on their meaning.

Karadžić 1.285 has the shortest introduction to the questions and answers of the five texts:

Later, in line nine, a fish in the water answers her soliloquy.

Mladenović 143 and 144, the two versions published by Karadžić, both open with five lines. The first has some kinship with Karadžić 1.285, which we have just seen:

In this case her questions are answered, not by a fish, but by an unnamed and hitherto unmentioned hero. Except for the last line, there is no verbal correspondence between this opening and that of Karadžić 285, but in both songs a girl is sitting by a body of water, a spring or the sea, and talking to herself. In 143 we do not know why she weeps; we are also puzzled by the sudden appearance of the riddle-answering hero. So also, it seems, was Karadžić, as he chose not to publish it. Mladenović 144 (in decasyllables) gives the girl some companions. {41|42}

In this text it is not a girl who asks the questions but an unmarried youth. His riddles were answered by Ana—who was sleeping! Probably it was this inconsistency that worried Karadžić and caused him not to publish the text. This opening has little in common with the others and seems to have been arbitrarily attached to the questions and answers. One cannot speak of textuality with reference to any of them.

The openings of Karadžić 1.286 and 5.379 are related to one another, but one is fuller than the other, twice as full to be exact. The former has nine lines, the latter eighteen. It turns out that both the riddlers and the answerers in each case are birds:

The first section of these songs (lines 1-2 of 1.286 and lines 1-12 of 5.379) is a bitter kind of lament, as it were, for a love that has passed, by a girl who has lost her lover. This section is elaborated in 379. In line 11 that variant turns its attention to what was evidently the lover’s complaint, that the girl has looked at someone else and embroidered a pillow for him. Even if she has, she says, she did not love him and she had embroidered a finer one for her lover. This is where the two versions meet, although in 286 the girl asks her lover if some other girl has given him an embroidered pillow. If so, she says that she has given him a finer one.

A degree of textuality, that is, of textual or verbal correspondence, actually begins at line 3 of 286 and at line 13 of 379, with the mention in both songs of an embroidered pillow, with the phrase vezen jagluk dala ‘given an embroidered pillow’ and with ja sam (ti ga) i ljepšega (dala) ‘I (gave) (you) an even finer one’ in 286 compared with ja sam (tebe) ljepši (ostavila) ‘I have (left) a finer one (for you)’ in 379. The descriptions of the figures embroidered on the pillows are somewhat different, but there are several elements and words common to both songs. In 286 there are thirty-three lions depicted, in the midst of which is a pavilion of coral, in which is an almond tree; in 379 there are thirty-three branches, in the midst of which is a branch of an almond tree. In both poems a bird is perched on a branch; in 286 it is a tica sevdelija, which I have translated as “a love-sick bird”; in 379 it is a tica mekušica, which {43|44} I have translated “a delicate bird,” No. 286 closes this section with the line Turski sjedi, a turski besjedi, “It sits in Turkish fashion and speaks in Turkish.” In 379 the bird on the branch is joined by a swallow, which comes from the sea and addresses the other bird with the questions that form the core of the song in all its variants.

In sum, then, this song has in the riddle section, as we saw, a more or less stable core, which exhibits a clear sense of textuality, including variables. It is preceded by a section (Karadžić 1.286, lines 3-8; 5.379,lines 13-17) that also has some degree of textuality, but the lines that lead into it vary considerably in the variants. We can speak here of the textuality of certain parts of the song.

With these five variants of a song in mind, let us now look at some other settings for the riddles. There is a group of them in which a lover awakes his beloved and asks the riddles; at least one of these also illustrates another aspect of the South Slavic women’s songs. From the appearance on the printed page of the songs in Karadžić’s collection (and that of others as well) one can form no idea of stanzaic shape or even whether they had any at all. Fortunately, in the Parry collection we have versions on the phonograph records of some of the songs in Karadžić. Although there is, of course, no way of telling whether the melody to which the songs are now sung was the one used in the nineteenth century, we can see what music the words were sung to in the twentieth. For example, Rabija Zvizdić in Gacko, Herzegovina, in 1935 sang a version of the riddle song in stanzas as follows:

The song is sung in stanzas consisting of couplets with a repetition of the last six syllables of the first line and the use of interjections, such as “Aj” and “aman,” at the beginning or the middle of a line. It is worth stressing that we are dealing for the most part, it would seem, with couplets. This is the same building block of composition that we saw in the Latvian dainas, and we will note it again and again as we proceed with further lyric material, also notably in English and Scottish ballads and even to some extent in South Slavic epic.

There are two other texts of this song in the Parry collection (Nos. 1577 and 2809) beginning with the same line, Dragi dragu alkatmerom budi, “A lover awoke his beloved with a red carnation.” They exhibit the same types of variation we have seen in the Karadžić variants. There are four texts in the collection (Nos. 9347, 9819, 10198, and 11176a) that begin with

and proceed with questions and answers as in the preceding texts. There are sixteen texts in the collection (Nos. 1294, 2264, 4040, 4227, 4645a, 7903, {45|46} 8641, 8667, 9157, 10289, 10453, 10490, 10511, 10548, 10580, and 12238) that begin with

and proceed as in other versions.

Other South Slavic settings for the riddles are more puzzling. Some of them place the girl by the sea or by a spring, and in them it is she who asks the riddles in a soliloquy. Sometimes she is weeping, and sometimes she is washing her face. In one song she is answered by a man, hitherto unmentioned. It is possible that this song is simply the result of confusion on the part of the singer. But there might perhaps be an implication that the girl is testing him to see what his caliber or intentions may be. Because, as remarked by Child, the Devil or another otherworldly figure sometimes appears in these songs, the testing of the man might not be inappropriate.

There are a number of separate songs that begin with Pod onom gorom zelenom, and it is in their context that we must try to evaluate our text. As we look at some of them, it is immediately apparent that the riddles are not found in them. The group that employ Vrani se konji igraju, “Black horses are prancing,” in an early line are of most interest to us, because that is the second line of our Gacko text. Karadžić 1.414 is an example:

Similar to this song are Karadžić 1.415 and 5.39. Except for the impatient horses and the weeping girl, these songs seem to have nothing to do with our riddles. An interesting variant is found in Mladenović 220:

This song adds a new scene to the preceding. Like Karadžić 5.39 it tells what happened to the fruit that was growing to maturity. Although the following song, Karadžić 1.291, does not belong in the Vrani se konji igraju, “Black horses are prancing,” group of the Pod onom gorom zelenom texts, it does have a relationship to the last section of the above.

Only the two Parry songs Pod onom gorom zelenom with which we began associate that beginning and the impatient horses with the riddle song. When the horses are described as watching a girl sitting beside the sea, the link is made with such songs as Đevojka sjedi kraj mora, “A girl is sitting beside the sea,” and we suddenly find her uttering riddles. Her first line, Ima li išta lepše od mene? “Is there anyone more beautiful than I?” leads her into a connection with a group of boasting songs, to which the riddles tenuously belong simply by the form of question and comparison. An example of such a song is Parry No. 4457 from Dzemila Pošković of Gacko:

One could undoubtedly discover links to other songs from this group of boasting songs, and so on from group to group in a seemingly never-ending network of associations. Thus songs modulate from one to another by means of associations of various kinds.

In the case of the Pod onom gorom zelenom songs adduced above, we can note not only the variations of the fruit parallel but also the addition to it of an account of what is done with the fruits, a normal progression of narrative. Finally, however, we saw a song in which flowers were treated in the same way as the fruits, but what was only a section of a song in one case is a completely independent song in the other. There is no green hill, there are no impatient horses, simply a girl picking flowers and offering them to father, mother, sister, and brother, all of whom refuse them. At last her lover accepts both the girl and the flowers. Here is a dramatic example of some of the ways in which songs are augmented and interrelated in tradition. The {55|56} riddles led us from “awaking” songs, to weeping maidens by bodies of water, to impatient horses, all of which are elements associated with marriage: choosing a bride (or bridegroom?) by means of riddles, maidens weeping because they do not wish to leave home, horses eager to set out to fetch the bride from her home. Tradition in the case of the lyric songs, at least, consists of a mass of interrelated and interwoven motifs and their formulaic expressions in discernible themes, or repeated passages, centered in the various stages of wedding ritual.

To return to the riddle songs, finally, in a class very much by themselves are Tica i djevojka, “The Bird and the Girl” (Karadžić 1.286) and Plač za dragijem, “Lament for One Beloved” (5.379) with their settings of a disappointed, and apparently unfaithful, girl trying to explain to her departed lover that she had always loved him more than the other man. To prove this, she cites the embroidered pillow she gave him, which was finer than the one she had given his rival. The pillow is described elaborately. On it, among other things, is a tree with branches on one of which sits a bird, representing the lover. He asks another bird the riddles, and she answers them, proving that she is worthy to marry him. In the other version the girl bird is sitting on the branch and a swallow from afar asks the riddles. Bird imagery is not uncommon in this poetic tradition, but the elaboration of the setting for the riddles is indeed unusual. The question of marriage is still the point of the riddling test.

We have seen that in some lyric songs in the South Slavic tradition there may be a section of the song at the beginning, in the middle, or at the end, which shows clear signs of a sense of textuality on the part of the singer, whereas other parts of the song may seem independent of textuality. There are some songs, of course, whose textuality embraces the whole song. I turn now to an example of such a song. Mladenović 316 has two variants, Karadžić 1.693 and 5.577. Here are the three texts:

Here is a humorous and satiric song, of which the South Slavic tradition has many, a parody, an inversion, of the usual love song. The young man complains that he was much sought for before he married but not desirable afterward.

In spite of the differences among the variants, including the fact that the first two are octosyllabic and the third decasyllabic, this whole song has textuality. Line six in Karadžić 1.693 is unique and breaks up the symmetry of introductory line plus the naming, which is repeated, of three female groups of acquaintances. In Mladenović 316 the singer has reversed lines four and five, also breaking the symmetry. It was probably for this reason that Karadžić did not publish the text. Only Karadžić 5.577 maintains the balanced structure and includes both udovice ‘widows’, otherwise only in 316, and nevjeste ‘brides’, otherwise only in 693.

The structure of this satiric song makes it easy to remember. The “before” and “after” lines are followed by a series of classes of women with a typical comment for each. It is remembered. There is no need to memorize the {57|58} song, because it can be put together readily enough from ordinary memory, and there is no fixed text to memorize anyway. The lines and half-lines of this song are made up of traditional formulas. The song, or the “theme” (that is, a repeated passage), is not improvised, nor is it memorized. Improvisation would mean that it was made up extempore of mainly nontraditional elements. The materials of the song are traditional, the structure is traditional, and it provides a plan that is easily followed. I believe that one might well say that the song is composed in performance. The stable opening of our first example of South Slavic lyric songs (Mladenović 155, Karadžić 1.590, 591) is also held together by its structure reinforced by such devices of sound as rhyme and alliteration. Its structure is a simple one of vocative plus question and answer. The form is that of a stylized dialogue. The alliteration of ružičice, ružo, and rumena is clear enough, as are the internal rhyme of djevojčice and ružičice and the end rhyme of rumena and medena. These are all structural elements that make both remembering and recomposing easy. So do the parallelisms in the riddle song. In the case of some lyric songs, then, composition in performance is the proper descriptive term for the process.

The Parry collection also has two dictated texts of this song from Đulsa Čampara (Nos. 4029 and 7323). They agree with Đula Hrustanović’s No. 12100 through the line Neka mu duša uzdiše! “Let his spirit sigh with longing!” No. 4029 ends there, but 7323 adds, in a different meter(!), lines suggested by the horse and obviously taken from another song:

In the case of both of these singers there is a stable core of the song and a tendency to go beyond that core by adding several more lines.

Đula Dizdarević dictated three versions of this little song to Čedika Šaković (No. 10528) and to Halid Dizdarević (Nos. 11740 and 11752). The first of these is like the first of Đula Hrustanović’s texts, ending with

No. 11740 begins with the two unexpected lines, then uses several familiar ones, and ends with new material, though an old subject:

Once again we have a more or less stable core with surrounding variations. No. 11752 has more surprises:

Although for the analysis of variants I have chosen songs I knew and liked and of which, from my work on the Bartók book, I was aware that there were a number of texts available, I believe that they are typical of their genre. In working with traditional songs and stories, it does not matter where one begins, because all songs and stories of a given traditional genre are interconnected.

An oral traditional song, no matter what its length or “genre,” textually, consists of one or more groups, large or small, of lines forming a more or less stable core adapted to context. Both the core and the lines used to adapt it consist typically of blocks of lines of varying size, but normally between two and four lines, linked in various ways that result (although probably not consciously conceived of for this purpose) in their being “memorable” and easily recalled when needed—or, more properly, when association brings them to mind. “Tension of essences” is close to “free association.” There seems, then, to be a principle operative in the oral traditional material considered which says that, no matter what the size of the composition, the traditional singers work with comparatively small blocks of lines intermediate between the formula and the theme.

Whether dealing with versions from a single singer or from several, one thus frequently encounters the more or less stable core of lines to which new elements are added. Sometimes they are added at the beginning. When they are placed after the “core,” it is almost as if after a given point in the song the singer were filling out the song with familiar traditional ideas and formulas. It can truly be said that even with short lyric songs, which are erroneously thought to have a fixed text kept verbatim in the memory, the concept of a memorized text needs to be modified. The larger the sample with which one works, the less adequate is the concept of word-for-word memorization as a means of song transmission.

Lord has demonstrated the oral composition of two large bodies of lyric poetry, Latvian dainas and Serbo-Croatian women’s songs. His method was to show that these short, nonnarrative poems were not memorized but were composed in performance by combining familiar, oft repeated blocks of lines (a stable, but not fixed, core of wording) with new or additional poetic {62|63} material. Although formulas abound in such poems, his method depended not so much on showing the proliferation of formulas as on demonstrating the persistence of blocks of lines, belonging to or leading to the compositional device of the “theme,” or repeated passage. He thus advanced from an analysis of short repeated phrases to larger units of thought.

The question naturally arises whether Parry and Lord’s methods can be applied to ancient Greek lyric and elegiac poetry in an effort to determine whether such poetry could have been composed orally. Milman Parry had indeed considered the question. In this “excursus” I gather together the pertinent remarks of Parry and other scholars. I quote them at some length in order to present their views as precisely as possible.