Many scholars (Kvideland, Ond, Lönnroth, Parry and others) have observed that variation is a given constituent of oral tradition – a way of thinking. Applied to traditional vocal music, this conception entails that there is no such thing as an original or correct version of a song. The essence of the notion is that the “work of art” comes into being the moment it is sung, which means that the existence of the song depends on our singing it. But if we create a song anew every time we sing it, and this is its only existence, where then does this thing we call the song begin and end – what is constant and what is variable?

In an attempt to shed some light on the phenomenon of variability, the present author has chosen to examine a set of songs as performed by seven different singers (born between 1860 and 1970). The study of variation is carried out by means of different methods of analysis, some of which are well proven and some are new in relation to the twin objective of both looking at variation in itself and assessing the applicability of the different methods. The article examines how the method of analysis influences the result. Methods such as western notation system, note-site analysis, melodic graph analysis, melodic skeleton analysis, syllable analysis, music-metric syllable raster, and sound analysis/acoustic analysis produce a degree of discrepant results or emphasise different parameters: tonal variation can be detected by using the western system of note writing, variation of form or variation of the melody-lines between different stanzas is best detected using melodic graphs, music metric syllable raster reveals variation in the number and placing of syllables in the phrases, sound analysis based on layered sound files yields information about tempo, phrasing and dynamics.