Individual Rhythmic Variation in Oral Poetry: The Runosong Performances of Seto Singers

The article explores the individual differences of rhythmic variation in traditional sung oral poetry. The analysed group of ritual songs is part of the Seto singing culture – a subtradition of Finnic oral trochaic tetrameter. The rhythmic variation, created by different positioning of stressed syllables in the song line, reveals itself on two levels, in linguistic verse structure and in musical performance. In the singers’ performances the verse and musical structures complement each other, having a cumulative rhythmic effect. By designing the rhythm at both levels, the singers systematically take into consideration the linguistic features of the used words. A statistical analysis shows a remarkable divergence in the rhythmic variation of different lead singers. The results are more homogeneous at the level of linguistic verse structure and more diverse at the level of musical performance. Also, the rhythmic choices of the lead singer and his or her choir in the course of the performance may differ. We may speculate that this divergence in individual rhythmic strategies could have been caused by the singers’ type of creativity and skills, their different perceptions of genre features and intergeneric relations, and the musical influences between more closely related singers.

styles (Oras 2004, Sarv 2008. Tracing the diachronic changes in the poetic creation of individual singers or the singers of different time periods is more complicated because of the formulaic nature of oral poetry -the texts of a certain moment are largely built on the language and formulas of the previous periods (Bailey 1993, Lotman 2011Frog, Sverdlov 2016: 10). At the same time, the sources of oral tradition (especially sound recordings, which are the most reliable source to study verse rhythm in songs) come from a rather short period of time.
In this article I will introduce the individual peculiarities in the rhythm of linguistic verse and in the realisation of the verses in the musical performance of traditional Seto singers. Seto singing tradition is one of the subtraditions of Finnic runosong, which remains in the southern border area of the runosong tradition. The main aim of my analysis is to consider the individual understanding of the poetic meter and verse rhythm by singers in actual performances. Hopefully such investigation will help to solve more common problems concerning historical developments of the runosong meter in Seto tradition. For the analysis I chose the sound recordings of a group of songs which show greater rhythmic variation than the Seto songs in general. The selected songs are a vivid and explicit example of how rhythmic variation is created in an oral poetic system. The source material also offers a good opportunity to demonstrate that, as in literary and art music tradition, in oral tradition as well, each individual creator shapes rhythm differently, combining the diverse devices of the linguistic, poetic and musical system. The obvious differences between the singers prompt the question of what may be the reasons and context for the individual rhythmic preferences. It is impossible to answer this question unambiguously. However, I will briefly touch upon the topic by questioning whether the similarities or dissimilarities between the performers could have stemmed from the mode of their creativity (whether they are more focused on the textual content or on the sound of the performance), the influences of other genres, the singers' connections to the older tradition, the family relations or creative ties between some singers, or the existence of local substyles.

Background
The base for the comparison of the rhythmic preferences of individual singers lies in the versification system of Seto songs in the context of Finnic runosong tradition, as well as peculiarities of the versification of analysed group of songs.
The runosong meter is the strictest in the northern region of Karelia, Viena. Here the polysyllabic word is positioned according its quantity: in the verse with eight positions and a trochaic rhythm, the uneven positions can be filled by any kind of syllables except short lexically stressed syllables, whereas the even positions can be filled by any kind of syllables except long lexically stressed syllables, and as an exception the first two verse positions can be filled freely (Frog, Sverdlov 2016: 8-9, Ross, Lehiste 2001: 2, 57-58, Sarv 2015. Significant linguistic changes have occurred ever since runosong developed in the Estonian language and also in the Seto language, which is linguistically considered a subdialect of the South Estonian language and has been traditionally looked at as a part of Võro dialect. The most important of these changes are shortening of the words (syncope and apocope), and development of the system of three quantity degrees to replace the former system of two quantity degrees.1 Owing to these changes the word stress has acquired more prominence and this has influenced the metrical system of traditional poetry.
1 In the Estonian language, the ternary opposition of quantity degrees has developed. There are three degrees of quantity: short (Q1), long (Q2), and overlong (Q3). The quantity distinction is a feature of the primary stressed disyllabic foot. The quantity of a word depends on the durational differences of the first (stressed) and second (unstressed) syllables, additionally also on their pitch contour, intensity and sound quality (e.g. Lippus et al. 2013, Asu et al. 2016. Concerning changes in the Estonian language see e.g., Viitso 2003, Sarv 2011a. Further on the subtraditions and changes in the versification system of runosong see Sarv 2015, Kallio 2017 Like the Estonian runosong, the Seto runosong could be described as a transitional form between two versification systems -the syllabic-accentual-quantitative (meter of Finnic oral poetry, runosong2, so-called Kalevala meter) and the syllabic-accentual system (Lotman 2009, Sarv 2011a, Oras, Iva 2017. In the Estonian and Seto tradition the syllable quantity does not determine the word positioning in the same way and to the same degree as in the Karelian runosong. Still, in most parts of Estonia, including the Seto region, word stresses have been preserved in the even positions in the verse and the words beginning in the even positions are mostly of short(er) quantity degree. In Estonian runosong, the words beginning in the even position are predominantly of the first quantity degree (Q1) which corresponds to the former short quantity degree. In the Seto runosong, the relationships and meaning of the quantity degrees seem to be changed: both the first and the second quantity degree occur in the function of the short quantity degree, and the third quantity degree contrasts with them and has a different metrical function (Oras, Iva 2017, cf. Sarv 2015.
At the same time, unlike the runosong of northern Karelia, the uneven positions in Estonian and Seto songs can be filled with syllables of any degree of quantity. The tendency to use Q1 initial syllables in the uneven positions is less visible in the songs of northern Estonia and more characteristic of the western and southern regions of Estonia, as well as of the Seto runosong. Since the stressed syllables of all lengths can be placed to the uneven position, there is no practical need to place short stressed syllables to the even position. Therefore, it seems that the former quantity model is gradually being replaced by the stress model: the aim of the singers is to occasionally fill the even positions with the stressed syllables for the purpose of achieving the accustomed rhythmic variety similar to the inherited older tradition rather than because of quantity (Oras, Iva 2017: 188, Sarv 2008. The rhythmic variations were observed in a group of ritual songs with refrain, where the linguistic verse rhythm is based on the common metrical system of Seto songs but is slightly more complex. Excluding the occasional or structural additional syllables at the beginning of the verses, it is possible to distinguish here between four classes of verses based on the positioning of stressed initial syllables of polysyllabic words in the metrical positions of a verse. 2) 8-position verses where the last stress of polysyllabic word is in the even position, two last stress groups have three positions (2+3+3 structure, x -/ x --/ x --; e.g. vee-me / kün-nel-de / kü-le-le, 'We'll take [him] to the candles', Fig. 2).
The most frequently used structure with the three-part stress group(s) is the second one (2). Here the word stress placement coincides with the most frequent broken line structure of runosong but does not entirely correspond to its quantity principles. The share of that structure among the analysed songs is 12,5%.
The other structures with the three-part stress group (3 and 4) are quite rare, occurring in 3.1% and 0.9% respectfully. In the context of current analysis, it is important that these two linguistic verse structures are performed only in one way without alternatives: to every syllable corresponds one isochronous note in a 7-note melody and the refrain immediately follows the last note without an exception. These structures could be considered as catalectic versions of the first two verse classes, but there are also reasons to interpret them as verses derived from the runosong verse structures 3+3+2 and 3+2+3 by shortening the first three-position stress group (Oras, Iva 2017: 184-185).

Material and method
The rhythmic variations were observed in a group of ritual songs with refrain, the melody of which belongs to an older layer of the singing tradition: harvest songs with lelo-lelo-lelo refrain (lelotamine), weddings songs with kas´ke-kan´ke refrain (kaaskõlõmine), and game songs with heiko-leiko refrain (leigotamine, 'The Horse Game'). The peculiarity of the older melody layer is the archaic one-three-semitone scale. As a rule, to each verse syllable corresponds a note of equal length in the melody -the syllabic verse rhythm precisely corresponds to the isochronous melody rhythm (Pärtlas 2001: 136-139, sound examples: ERmA no. 10, 50). I have analysed 25 recordings of the harvest song (419 verses), 26 recordings of the wedding song (282 verses), 11 recordings of the game song (449 verses). All 62 recordings, with 1150 verses in total, belong to    I examined the shaping of the rhythmic variety at two levels: at the level of linguistic verse design and at the level of the musical performance of the verse. Analysing the oral sung poetry, it is possible to differentiate between the linguistic verse rhythm and the rhythm of a (musical) performance, but in many cases we need to know the actual performance tradition to understand the principles of the placement of the linguistic units into verse positions (Sarv 2011b: 332-334;cf. Leino 1986: 141-143, Kiparsky 2006, Dell 2015, Frog, Sverdlov 2016, Niemi 2016.
In my study, I asked about these principles, relying on the sound recordings of local performance tradition. I analysed the placement of stressed syllables of polysyllabic words into verse positions, basing on the fact that in runosong the musical units, usually sung with one text syllable, correspond to verse positions. The cases of performing linguistically identical verse structures with a different musical rhythm I observed as a next rhythmic level -the rhythm of musical performance. The differentiation helped to outline the differences in the preferences of individual performers: whether the singer prefers to use the more "complex" variation of the linguistic verse rhythm, i.e. variation at a deeper structural level, or s/he prefers the "simpler" variation of the rhythm, i.e. the use of alternative rhythmic solutions in the musical performance of the linguistically same verse.
Variation of the linguistic verse rhythm is based on the use of the four classes of linguistic verse structures introduced above. The verses of the first class containing only two-position stress groups model the metrical norm. Hereafter I will call them 'even verses'. The task of other verse classes (classes 2-4) is to create the rhythmic variation by adding three-position stress groups to the "customary" two-position stress groups. Hereafter these will be called 'substantial uneven verses'.
The possibilities of using rhythmic variation during the performance, however, are not limited to the use of the substantial uneven verses. Some even verses (verses that belong to the first class), can be interpreted in performance with the same musical rhythm as the verses of the fourth verse class. In that case they are sung with seven isochronous notes, whereas the second half of the verse is performed shorter than usual: the syllables which can be (and often are) sung with four notes are "compressed" to three notes so that it sounds rhythmically like the verse structure 2+2+3 (i.e. verse class 4). To "compress" the second half of the verse in a musical performance to three isochronous notes instead of four notes, (a) the first two syllables of a Q1 or Q2 word are sung half shorter (Fig. 5); or (b) one Q3 syllable is sung half shorter, with isochronous one note instead of the usual two notes (Fig. 6). The method of shortening is connected to the metrical pattern characteristic of Seto runosongs and the songs of the neighbouring Võru district in South-Estonia: usually the initial syllables of the Q3 trisyllabic word fill two verse positions and are normally sung with two isochronous notes (one double-long note). (In the local dialect the Q3 syllable -unlike the Q1 and Q2 syllables -builds a separate foot in the local language.) Hereafter I will refer to these "compressed" musical lines as 'shortened lines'.
During song performance, the singers can use two possibilities of creating a musical rhythm different from the rhythm of the most common even musical lines: (1) they create substantial uneven verses (verse classes 2-4, Fig. 2-4) and (2) they shorten the second half of the verses with even linguistic verse structure in the musical performance (verse class 1, Fig. 5b, 6b). The first possibility is in principle "more complex" because the rhythm of the song is varied at the level of linguistic verse structure. The other possibility, shortening, can be understood as a "simpler", secondary method to create the musical rhythm identical to the musical rhythm of substantial uneven verses in performance.
While analysing the material I identified the musical uneven lines among all musical lines, considering separately the cases of "complex" and "simpler" rhythmic variation. On the basis of these data I compared statistically the individual rhythmic choices of the lead singers, from whom at least 20 song lines have been recorded.
The exciting additional nuance of varying rhythm is that the lead singer and the choir may not necessarily assign the same rhythmic interpretation to the same verses. The role of the lead singer is to sing the verses first alone and the choir repeats the verse sung out by the lead singer. Generally, the choir includes less than ten singers. The possible "disagreement" in the choice of rhythm concerns the linguistic verses with the even, symmetrical stress structure (verse class 1): the lead singer may use the shortening of the line, so that the result would be a combination of musical metrical groups of 2-2-3 structure, while the choir remains true to the symmetrical 2-2-2-2 structure; sometimes it is the other way around.
The cases of "disagreement" of the lead singer and the choir were determined based on hearing, excluding the most obscure variants (the cases where the second half of the verse is performed in a way that its duration seems to remain between three and four isochronous notes). Among the shortened lines, I also included these not-so-rare cases in which the shortening is performed only by some members of the choir, so that there is variation within the choir. The differences in the rhythmic interpretation of the lead singer and the choir in the corpus of recordings of lead singers with 20 or more recorded song lines were observed. I calculated the percentage of verses interpreted differently by the lead singer and the choir and inquired as to which of the two possible interpretations either of them prefers.

Results
The statistical comparison of the data of 17 singers showed similarities but also remarkable differences in their rhythmic preferences (Fig. 7). The average percentage of all uneven musical lines (emerged both in the course of performing of the substantial uneven verses and as the result of the shortening of linguistic even verses in a performance) is 44% (with a standard deviation of 20). More than half of the singers fall in the range of 33-47%. The percentage of all uneven musical lines ranging above the average is achieved through an extensive use of the "simpler" means -shortening linguistic even verses in a performance. The percentage of the substantial uneven verses by almost all the singers is less than 30%, whereas such verses are not used by some singers. The average of substantial uneven verses is 18% (with a standard deviation of 12). A remarkable exception here is one of the most famous Seto singers of all times Anne Vabarna (1877-1964) and her daughter Evdokia (Ode) Vabarna (the number of lines recorded from them is not large, consisting in 34 and 29 lines, respectively). The sample from both singers include 41% of substantial uneven verses. While Anne Vabarna's repertoire did not include any shortened lines, the sample recorded from Evdokia Vabarna includes quite many, resulting in the largest percentage of all uneven musical lines (83%) in her performance among all other singers.
If we focus on the singers with a slightly more reliable sample, that is the six singers from whom 60 or more lines have been recorded, the results are remarkably more homogeneous (Fig. 8). The samples of all these singers contain uneven musical lines of either kind, substantial uneven verses as well as shortened lines. Still, there is a considerable difference between the singers with the lowest and the highest percentage of uneven musical lines, Akulina Pihla and Matryona Aalik, with 25% and 74% of song lines, respectively. The percentage of the remaining four singers remains between 37% and 44%. As I have pointed out in the case of the larger group of singers (from whom 20 and more lines have been recorded), the larger percentage of uneven lines is generally achieved by means of shortening -for example, Matryona Aalik has used shortening of the linguistic even verses in 52% of all of her performed lines. The percentage of shortened lines used by the remaining five singers is remarkably smaller, falling in the range of 12%-25%. The percentage of substantial uneven verses in the sample of all six singers is 13%-24%.
So far the results have concerned only data on the verse and performance rhythm of the lead singers. An analysis of the cases of different rhythmic interpretation of linguistic even verses by the lead singer and her/s choir showed that disagreements in the interpretation occur in the case of 12 singers out of 17. The lead singer and the choir may sing the same verses with a different musical rhythm (e.g., one of them uses shortening and another does not) in up to half of all the shortened lines. In all the instances of shortening, the percentage of such disagreement by six singers and their choir is between 14% and 41%. In the cases of disagreement, the method of shortening is mostly used by the lead singer -her/s choir repeats the verse without shortening, i.e. with the ordinary even musical rhythm in a 8-note line. Still there are a few cases in which the choir opposes the variant with an even musical rhythm chosen by the lead singer and decides to use shortening. This occurs more than once in the cases of two choirs. For example, the choir (or part of the choir) of Akulina Pihla has used, unlike the lead singer Pihla herself, shortening on six occasions altogether, whereas the reverse -Akulina Pihla singing a shortened line and the choir repeating the line with the ordinary even musical rhythm -happens only once.

Discussion
The musical rhythm of Seto refrain songs in the old style is seemingly simple: in the isochronous rhythm, the verse positions of trochaic syllabic verse correspond to the eight notes of equal length. Nevertheless, the rhythm is highly diverse due to the possibility of creating three-position word stress groups and combining them with the two-position stress groups. In music, metrical groups of three isochronous notes correspond to the linguistic three-position stress groups. In songs with refrain there are three linguistic verse structures with three-position stress groups, in two of them the number of syllables/positions/notes is not eight but seven. The singers seemed to need the rhythmic variety created by these structures -on average they use an uneven stress group structure in one-fifth of all the verses (in the current analysis such verses are called substantial uneven verses). However, the singers did not seem to be content with using only this kind of rhythmic variation. In addition, some linguistic even verses are also contracted in singing by means of shortening, so that in the performance sounds a musical three-part metric group (2-2-3). I would suppose that the means of shortening has been inspired by substantial uneven verses -and the latter, in turn, by the verses where the lexically stressed syllables fall to the metrically unstressed position (also called "broken verses") of typical Finnic runosong.
The number of shortened lines reflects most clearly the individuality of singers. Some singers used them so often that there are more of those than even musical lines, while one singer did not use them at all. There are singers for whom the method of shortening is the only way to shape uneven musical linesthese singers avoid the substantial uneven verses completely. So the shortening functions not only as an additional possibility to vary rhythm but also as an alternative to creating linguistic uneven verse structures. It is possible that for some singers such means may have been a simpler and more convenient way to create the rhythmic variety they had been accustomed to and needed. What could be the reason for such remarkable differences in the use of uneven lines by the singers, or the extreme cases in which some singers avoided either the substantial uneven verses or shortened lines? Quite an obvious problem concerning the source material and one of the reasons for the major differences in the data is the relatively small number of song lines recorded from a singer. It is plausible that if more song lines were recorded from each singer the overall picture would be more homogeneous, at least to a certain extent -as it is, for example, the case with the data on six singers from whom a larger number of lines have been recorded. Statistically more reliable is the data on the six singers, which shows that although most of them remained within the average number of uneven musical lines, they still used them to a varying degree. There is a remarkable difference between the lead singer Akulina Pihla who used rhythm variation the least and Matryona Aalik who used more uneven lines in her musical performance than any other singer.
One could assume that a more objective and common reason for the differences between individual singers is perhaps regional peculiarities, which is typical to any kind of variation in historical ethnographic data. The results, however, do not reveal that the singers of certain regions shared similar characteristics, quite the opposite -singers from the same region may yield radically different data. Thus, it is most reasonable to prefer the idea about the individual perceptions of the performers and their personal creative style (incl. use of rhythmic variation).
Each singer may have perceived the relations of the analysed song group with other song genres differently. An attempt to interpret the rhythmic structures of the songs analysed here as the structures derived from trochaic tetrameter of Finnic runosong is a tempting, but far too simplistic, solution. It must be remembered that the songs without refrain in the old style contain only linguistic uneven verses (and correspondingly also musical lines) of the stress structure 2+3+3, and the 7-position structures are impossible in these (Oras, Iva 2017). On the other hand, one genre in the Seto collective singing culture -namely, choral laments -shows even greater rhythmic variation, not to mention the solo laments, which show the greatest rhythmic freedom (Pärtlas 2001: 137, Ross, Lehiste 2001: 124-125, Sarv 2000.4 The rhythm of both solo and choral laments cannot be directly linked to the Finnic runosong. Perhaps the variation of rhythm, and the extent to which different singers need the variation, depends on whether the singers perceive the rhythmic structures of the analysed songs as being more similar to the songs with lines of equal length or to the rhythmic model of choral laments.
There were some famous and colourful personalities among the singers, the data on whom is somehow conspicuous or exceptional. For example, Akulina Pihla's performances reveal her modest inner motivation to use "special" rhythms. Quite telling is the reaction of Akulina Pihla's choir to her lead singing -she was the lead singer whose verses, sung with even rhythm by herself, were in many cases interpreted by the choir as uneven musical lines, using the means of shortening. Apparently someone in Akulina Pihla's choir felt the need for a richer rhythmic variation than Pihla chose to use. A traditional choir was subject to a hierarchy in which the lead singers who are very knowledgeable in the tradition, as well the singers of upper solo voice in the choir, occupied the higher position. It is possible that among the members of Akulina Pihla's choir there were also other lead singers or dominant personalities who contrasted their singing with Pihla's singing using traditional rhythmic devices in order to establish their own creative style. Such a situation could be compared with the accounts from the Kihnu Island off the western coast of Estonia. On the Kihnu Island, runosongs are sung together but one of the singers assumes the leader's position, determining which verse or text motive will be sung next. According to musicologist Udo Kolk, a 4 Relying on the studies of Leonard B. Meyer, Reuven Tsur has distinguished between the different emotional meaning of the convergent and divergent style of poetry: the divergent style yields increasingly emotional and sometimes witty poetry, the convergent style creates the feeling of "simplified mastery of reality, as in nursery rhymes and folk poetry" (Tsur 2017). Actually, the divergent style can also exist in folk poetry, of which the Seto choir laments and also the analysed group of songs are good examples. It is meaningful that the analysed songs are ritual songs sung during the liminal stages in the life of the community. The instability of liminal stages (Turner 1969) seems to be connected with the rhythmical divergence of the performance -the instability of a situation may not only to be in accord with emotionality characteristic of the divergent style, but it also can be managed by this divergent style: the rhythmical "disorder" can be seen as a device to bedevil and thus keep away the irrational damaging powers, like many other magical practices during rituals. meticulous folklore collector, a singer who aspired to become a leader of the choir would start to sing more actively, introducing simultaneously some modifications to the melody (Kõiva 1964(Kõiva : 8-9, 1965. Akulina Pihla, the long-term lead singer of the Leiko choir in Värska, was widely known throughout the Seto region as a "mother of song" (traditionally the most valued lead singer, e.g. Kuutma 2006: 136). The Seto community particularly appreciated her excellent knowledge of song lyrics and her ability to improvise striking lyrics for every occasion (Kalkun 2008). I believe that the performers indeed employed different tools to realise their creativity. Some performers intuitively might have contributed more to a musically versatile performance, while others were more focused on other aspects of singing. Evidently, Akulina Pihla focused more on the content and message of the songs, whereas the metrical and musical design of rhythm occupied a secondary place in her creation.
Anne Vabarna, one of the greatest Seto lead singers, stood out among others with her particularly large number of substantial uneven verses and the absence of shortened lines. The percentage of substantial uneven lines in the singing of her daughter Evdokia Vabarna was the same, in addition she used many shortened lines. Although the number of song lines recorded from both singers is not very large, these cannot be ignored because Anne Vabarna was an exceptional specialist of the older tradition and also a masterful improviser (Kuutma 2006: 140-149). It is plausible that her highly frequent use of more "complex" rhythmic variants shows her good knowledge of the tradition and skill in creating song texts. Furthermore, one might argue that she did not use the "simple" means of shortening because she was able to achieve the expectable (average) musical rhythmic variation already at the level of linguistic verse structure, with more "complex" means. Anne Vabarna's daughter Evdokia, who additionally employed the means of shortening, seemed to use uneven rhythm in her performance rather excessively and it is difficult to speculate why. At the same time, her rhythm design principles are not too exceptional, as there are some other singers who also used a considerably small share of melody lines with an even rhythm in their performances.
In the light of Anne Vabarna's principles of rhythmic variation, it is worth pointing out the relatively large share of substantial uneven verses in the singing of Elena Laanetu -the percentage of these verses in her singing is the largest among the six singers from whom more recordings were made. Elena Laanetu was a relative of Anne Vabarna and had sung with her on many occasions (Sarv 1997: 180). The large number of substantial uneven verses could perhaps be explained by the influence that Anne Vabarna, a singer of the older generation, exerted on younger Elena Laanetu when they sang together. Still, the share of substantial uneven verses in the singing of Elena Laanetu is by no means exceptional. Some other singers have almost as many of these verses, among them Matryona Aalik, who sang in the same choir with Elena Laanetu. The remarkably different number of shortened lines used by Elena Laanetu and Matryona Aalik could be seen as a telling example of the differences of women singing in the same choir -while the percentage of shortened lines used by Laanetu is one of the smallest among all the singers, that used by Aalik is one of the largest.

Conclusion
The present article focused on practical interpretations of the verse meter by individual singers, taking into account both the linguistic structures and their realisation in a musical performance. Detecting the singers' interpretation of the local metrical system may lead to a proper understanding of the meter of the Seto traditional songs, opening new possibilities also for purely linguistic or metrical analyses.
The analysis shows remarkable differences between the rhythmic principles of individual singers. Analysing these differences separately on two levels -the level of linguistic verse structure and the level of musical performance -it turned out that at the level of linguistic verse structure the data on rhythmic variability was considerably more homogeneous than at the level of musical performance. The number of linguistic structures with uneven placement of word stresses (i.e. structures consisting of 3-part and 2-part stress groups) diverging from normative, even structures (with only 2-part stress groups) was moderate. Still, some differences in the rhythmic preferences of individual singers could also be detected at the level of verse structure. At the same time, the number of so-called shortened lines which emerged in a musical performance of linguistic even structures varied considerably from one singer to another and highlighted more prominently their individuality in conceptions of rhythm. One could even speculate that some singers employed "simpler", musical means of varying the rhythm as an alternative to the more "complex" production of uneven linguistic structures (called substantial uneven verses here). The singers themselves supposedly did not consciously differentiate between these two levels -their aim seems to be just creating rhythmic variation.
Explaining the different rhythmic preferences of the singers with regional peculiarities did not prove successful. It does not seem to be possible to tie the diverging preferences of the singers to a specific historical period in oral tradition. (The exceptional abundance of linguistic uneven verse structures and the lack of musically shortened lines of the excellent singer Anne Vabarna could be connected to her good knowledge of the older tradition and her early birth year, although this assumption is hard to prove.) The differences and similarities can be explained rather by individual preferences which may be possibly influenced by the nature of creativity and skills of the singers, different perceptions of genre features and intergeneric relations, and the personal relationships and influences between the singers.