Arthur Holmer
Search for other papers by Arthur Holmer in
Current site
Google Scholar
Håkan Lundström
Search for other papers by Håkan Lundström in
Current site
Google Scholar
Seediq canonic imitation (Taiwan)

The chapter deals with Seediq vocal expressions that start with a short motif that is often repeated once and then followed by another motif, which is repeated once etc. This pattern is sometimes called ‘litany form’. The form that results when two persons sing in alternation is called canonic repetition. It is shown that such performances can be understood as being built on performance templates that enable singers to vary their words.

There are several ethnic groups in Taiwan belonging to the Austronesian language family. Their musical traditions are characterized by various forms of canon performances, which are often polyphonic. There is a high degree of improvisation, and the main performer will be followed by the rest of the participants when they recognize the words or when there are recurring refrains. Vocables – or non-lexical words – are very common. They may occur in combination with lexical words, but entire performances may be made up of vocables that are rather fixed in each case.1

The Seediq are one of the most recent groups to be officially recognized as one of Taiwan’s indigenous groups, in 2008. They were previously classified as a sub-group of the Atayal, together with the closely related Truku or Taroko people, who were recognized as an independent tribe in 2004 (Map 4). The Seediq language is spoken to the north-east of Puli in Central Taiwan, past Wushe and the valleys to the east and north-east, and across the Central Range to the Taroko Gorge and the Pacific coast. It belongs to the Atayalic group of languages, which, in addition to Seediq, comprises Squliq and Ci’uli (both often referred to as dialects of Atayal). Linguistically, Seediq is divided into three dialects: Truku or Taroko, spoken in Hualien County (and today officially recognized as the separate language of the independent Taroko tribe); Toda, spoken in a couple of villages in the north-eastern parts of Nantou County; and Tgdaya, spoken between Puli and Wushe, and in and around Wushe itself. The data analysed here are from the Tgdaya dialect as spoken in Chingliu, and Chungyuan in the Guoxing valley north of Puli.

An example of Seediq canonical singing is included in the movie Warriors of the Rainbow: Seediq Bale, released in 2011. It is a historical drama built on the Wushe Rebellion, which began in October 1930 and formed an uprising against the Japanese forces in Japanese Taiwan.2

The Seediq language and vocal expressions

Though there are some studies of the music of ethnic groups in Taiwan, and some of those incorporate linguistic aspects, there are very few descriptions of the music of Seediq. Our main objective has been to study vocal expressions of the form that contains repetitions of phrases, in order to see to what extent the concept of the performance template is relevant as an approach to analysis. This was done by transcribing the verbal phrases performed into notation, and by investigating relationships between them.

Two Seediq performances will be discussed here. They are similar, forming dialogues about a maiden and a man meeting in the forest to go together and work in the swidden fields. Two females perform them together: Uma Watan and Lubi Mahung, both from the village of Gluban (Chingliu 清流 in Mandarin). Translation is complicated by the fact that some words are archaic, and by the presence of vocables – the most common ones being wa, e, ewa, siyo. The exact boundary between what is a vocable and what is an archaism may be hard to establish. One possible criterion might be that if an expression does not appear to obey the phonotactic patterns of Seediq, it can safely be viewed as a vocable; but, given that phonotactic patterns may change, this is not entirely reliable either. One such phonotactic pattern concerns vowel realization. In Tgdaya Seediq, stress is on the penultimate syllable. Vowels preceding the stressed syllable are systematically elided, and the resulting underlying sequences of consonants are re-syllabified with epenthetic vowels where necessary. In speech, these epenthetic vowels are usually realized as [u], although this is not always very clear, as they are unstressed. In vocal expressions, however, they may surface in a stressed position, because the stress patterns of the lyrics and the melody do not necessarily match. When this happens, the vowel quality [u] is very clear. Furthermore, cases like these show that the morphophonemic changes in Seediq are not just an automatic question of the realization, but have become phonological.

For both performances, very rough translations of the lexical items or phrases that are clear have been added. This serves to give a general picture of the meaning. More detailed surveys of the lyrics become quite difficult, given the preponderance of archaisms and vocables mixed with the lyrics, as well as single interspersed lexical items.

Analysis 19 Uuyas obio, ‘The obio song’

Obio is a vocable, and uuyas is a manner of performing a vocal expression that is usually translated as ‘song’. This is a dialogue about a maiden and a man meeting in the forest and going to work together in the swidden fields. It has five sections (Example 98 A–E) which consist of phrases performed to short musical motifs. There are two persons who perform in unison, and each phrase is repeated once (in the transcription of the words, though, the poetic phrases are written out only once, while the first phrase is written out with the repetition in the notated example that starts each section):

o- yo- su na o- yos             o- yo- su na o- yos …

The repetition is either identical or with slight changes in pitch, particularly on the final tone, so that the final tone is lower in the repetition. The effect is that of an echo, and musically the second part mirrors the first so that the two phrases sound like a unit. For each section (A–E), the notated introductory unit is repeated throughout, with minor variations.

Each section functions as a stanza held together by sound rhymes (vowel rhymes, consonant rhymes), repetition, and parallelism. Each stanza has its own metre. Exclamations (section A) and questions (section B) parallel spoken language fairly clearly, and speech rhythm and intonation may play important parts in the rhythmic and tonal characteristics of the corresponding musical motifs.

In the following, the lyrics will be presented, as far as they can be identified, without differentiating between actual words, vocables, and possible archaisms which we have not identified as actual words. Epenthetic vowels that are added in speech but are almost inaudible are clearly audible in performance, and they are written in italics (normally the vowel u). Rough translations are provided, not so much to give the exact meaning of each line, but to convey the general meaning of the entire stanza. It follows that the lyrics have not been glossed.

Example 98 A–E Obio. Original pitch: f# ≈ 203 Hz. 21 Obio

Section A (0:0)

9 syllables: (2+1+)5+1 (ᴗ ᴗ ––) ᴗ – ᴗ – ᴗ ––

o- bi-yo!
ya- hi ku sumba- rux
e- wa de musa- re
sa- re yo sa- re
o- bi-yo!
ya- hi ku sunter- ung
ce- ka be qute- lun
sa- re yo sa- re

Obio! Come to work the fields with me. … maiden …

Obio! Come and meet me

in the middle of the forest path …

Section B (0:26)

6 syllables: 5+1 ᴗ – ᴗ – ᴗ ––

o- yosu na o- yos
i- ma si mae- kan
pa diyax su- nara ki
o- yosu na o- yos

Who will eat?

Section C (0:48)

5 syllables: 4+1 ᴗ – ᴗ – ––

i- ma ngayan su?
nga- yan mu I- buy
i- ma su i- su?
mu- lawa ba- kan
i- ma ku ya- ku?
Nu- harung Ba- so

what is your name?

I am Ibuy, I am Ibuy, Who are you?

bakan calls

who am I

Harung, son of Baso (or possibly Baso is Harung’s)

Section D (1:19)

6 syllables: 5+1 ᴗ – ᴗ – ᴗ ––

li- xo rumading wa
e- wa kisoring ni
tu- lame texan ho
n Tado Robo ni

let us leave off beginning

let us try once

Robo is Tado’s

Section E (1:39)

7 syllables: 2+1+4 ᴗ ᴗ –– ᴗ ᴗ ᴗ

suwa- e ta da- ki ni
n-ni- ta nii da ho

farewell, let us part

Analysis 20 Meeting and working in the fields

In this performance, the content resembles Uuyas obio, and some of the words are the same or similar. It is performed in the form of canonic imitation, a way of constructing music that is well known in what is possibly its most basic form in Seediq practice.3 The canonic imitation works like this: the lead performer starts a phrase and repeats it just as in the previous performance, whereupon the second performer repeats the same phrase twice, with a delay. In this case the delay is generally arranged so that the main beats coincide, but in some longer phrases the rhythmic relationship is more complex. In principle, then, each performer repeats each phrase twice, i.e. it is performed four times before the next phrase starts:

e-  nu-  gu  a-  gu-  uh  e-  nu-  gu  a-  gu-  uh …

       e-  nu-  gu  a-  gu-  uh.  e-  nu-  gu  a-  gu-  uh …

There are seven sections (Example 99 F–L). Each section has its own motif which is coordinated with the number of syllables in the line and with different distributions of stressed and unstressed syllables. A very similar construction of a ‘Tayal song’ (i.e. a performance from the closely related Atayal group) has six sections with similar rhythmic motifs.4 A further similarity is that longer lines are performed slowly, resulting in more complex rhythmical relationships between the parts of the two performers.

Example 99 F–L Meeting and working in the fields. Original pitch; f# 185 Hz. 22 Working in the fields

Section F (0:0)

5 syllables: 4+1 ᴗ – ᴗ ᴗ —

e- nugu agu- uh
hai- ta ru- meno- o
e wa mu- sare- e
tu- lame texa- an
sa malu exa- an
ka pera- dinga- an
wa kesing na we- e

Let us play round, maiden*

let us try once,

the beginning

Note: * Msare is glossed by consultants as meaning the same as weewa ‘maiden, young woman’. It is actually attested only in song lyrics, however, and only as a form of address, or possibly an exclamation, suggesting that it is not necessarily a form that could be used syntactically as a noun.

Section G (0:35)

5 syllables: 1+3+1 — ᴗ – ᴗ —

e- e ka’- i be- nah
na lebi i haan ta

Section H (0:46)

6 syllables: 5+1 ᴗ – ᴗ – ᴗ —

li- xo ru- mading wa
e- wa mu- sare ni
te- lame texan ho
na pera- ding han ho
e- wa mu- sare ni

let us leave off beginning


let us try once

the beginning

Section I (1:10)

4 syllables: 2+1+1 ᴗ ᴗ — —

n-na beno   mi

5 syllables: 3+1+1 ᴗ ᴗ ᴗ — —

sino qu-leng- un
sino ma Te- mi
hai be tuma- ra
ceka qu-te- lun
e n nami nii

wine is necessary

wine and Temi*

go really and wait

in the middle of the forest path

Note: * This line may have been added spontaneously as a joke, since one of the listeners present is called Temi.

Section J (1:46)

6 syllables: 2+2+2 – ᴗ – ᴗ – ᴗ

a sumpu* sum- barux
a kedi ku- hayan
a riso su- lawa
tu-lame ta texan
o nawe ta- wa so

work in the fields together

young man,

let us try once

Note: * This word is sung as [sumpu] in this recording but is presumably intended as suupu ‘together’, in which case it occupies the correct position in the phrase suupu sumbarux ‘work in the fields together’.

Section K (2:13)

5 syllables: 4+1 ᴗ ᴗ ᴗ ᴗ —

siyo, siyo, si-i[yo]
rumeno wa so-o
mutara hi- ya-a
ewa musa- re-e

play around

wait there


Section L (2:34)

6 syllables: 5+1 ᴗ – ᴗ – ᴗ –

nio tu di tan- da
lixo ru- mading wa
ewa ki so ring- wi
tulame texan ho
nu Tado Rabo ni

let us leave off beginning

let us try once

Rabo is Tado’s

Poetry, metre, and rhythm

Generally speaking, a section forms a stanza and a stanza stands out as a poetic unit based on parallelism, repetitions, and rhymes. In the following discussion, references are made to Analysis 19 (Example 98, sections A–E) and Analysis 20 (Example 99, sections F–L). In Analysis 19 section A, two lines are repeated (obio and sare yo sare), a beginning of a line is repeated (yahi ku…), there are rhymes built on similarity of sound (sumbarux / sunterung / qutelu) and final rhyme (musare / sare), which can also be seen as a chain rhyme where one of the last words in a line rhymes with one of the first in the following line (musare / sare yo). Vowel and consonant rhymes are also numerous.


yahi ku sumbarux

ewa de musare,

sare yo sare


yahi ku sunterung

ceka be qutelun

sare yo sare

Similar techniques are found, for instance, in sections B and H for repeated lines, and in C and J for repeated words at the beginning of lines. In other stanzas, a final rhyme or a vowel rhyme dominates. In this manner the stanzas are firmly tied together while, in turn, they sometimes have their own distinctive metres. The dominating poetic metres of the two performances are listed in Table 7. There are some variations among them, and there are a few cases of lines prolonged with one syllable within a stanza. In such a case, an extra syllable is then added to the short ones in the metre within the same time slot.

Lines of 5–6 syllables are the most common, and the iambic pattern dominates. The most common way to perform an iambic 6-syllable line starts with an upbeat: - | – - – - –––. A 5-syllable line may just tie the two last tones together: - | – - – -–––. In this rhythmic pattern, it is essential that the epenthetic vowel [u], which otherwise occurs in unstressed positions only, is given the same length as other vowels, and this differs from speech. Interestingly, the 5-syllable line (C) has a different pattern: ᴗ – ᴗ – –– realized - | - - - –––. This may be because most of the words themselves are fairly short, and they do not (with the exception of m[u]-lawa and n[u]-harung) contain any underlying consonant clusters which could be syllabified into pre-stress vowels. Alternatively, given that several lines in (C) are questions, it might reflect speech intonation.

Section Words Syllables Performed Metre Rhythm
A yahi ku sumbarux 6 5+1 ᴗ – ᴗ – ᴗ –– - | – - – - ––
B oyosu na oyos 6 5+1 ᴗ – ᴗ – ᴗ –– - | – - – - ––
D lixo rumading wa 6 5+1 ᴗ – ᴗ – ᴗ –– - | – - – - ––
G e-e ka’i benah 6 5+1 ᴗ – ᴗ – ᴗ — - | – - – - ––
H lixo rumading wa 6 5+1 ᴗ – ᴗ – ᴗ — - | – - – - ––
L nio tu di tanda 6 5+1 ᴗ – ᴗ – ᴗ – - | – - – - –
F enugu aguuh 5 4+1 ᴗ – ᴗ ᴗ — - | – - – -–––
C ima ngayan su? 5 4+1 ᴗ – ᴗ – –– - | - - - –––
J a sumpu sumbarux 6 6 – ᴗ – ᴗ – ᴗ – - – - – -
K siyo, siyo, si 5 4+1 ᴗᴗᴗᴗ ––– – - – - –––
A obio 3 2+1 ᴗ ᴗ –– - - | –––
E suwae tadaki ni 7 2+1+4 ᴗ ᴗ –– ᴗ ᴗ ᴗ ᴗ - - | ––– – - – -
I sino qulengun 5 3+2 ᴗ ᴗ ᴗ — — - - ––– | –––

Trochee patterns (J, K) are musically performed in the same rhythm by discarding the upbeat; 6 syllables: – - – - – - , and 5 syllables: – - – - –––.

There are some sections that begin with an anapest: - - | –––. One of them is the vocable obio (A), while others are longer. These sections are performed more slowly, and the stresses fall differently. One of them was transcribed as 5/4 beat rhythm, but whether these lines should be interpreted as an instance of triple metre (3/4) might also be discussed. These sections stand out in the performance by having a contrasting slow rhythm, compared to the generally fast 4-beat delivery that dominates the other sectors. It is also connected with words that are emotional in content, like the pleading obio! come to work the fields with me. … maiden … (A) and go really and wait, in the middle of the forest path (I). It also occurs in the final sections of the first performance, which is a farewell: farewell, let us part (E).

In performance, then, the tightly woven poetry of the stanzas is combined with strict metres and rather fixed rhythmical patterns. While this sets up a number of criteria that open the field for improvisation of new words, it would also make existing lines or stanzas more stable. This may explain the use of unchanged vocables as well as the use of archaic words which may have lost most or all of their meaning.

The Seediq imitation performance template


  • Built on short melodic motifs of less than 4 beats, with a tonal centre (see Examples 98, 99).
  • Speech rhythm and intonation affect the tonal realization.
  • Range: 7 semitones.


  • Regular pulse.
  • Speech rhythm affects the rhythmic realization.
  • Prosodic phrases are performed to variants of: - | – - – - ––, where the first tone in the bar and the last long one are stressed.


  • Litany: dialogue divided into sections or stanzas.
  • A stanza is held together by sound rhymes (vowel rhymes, consonant rhymes), repetition, and parallelism.
  • A stanza starts with a line that has a different number of syllables from the preceding one, as well as a new variant of the short rhythmic/melodic motif.
  • In imitation style, the performer (or performers in unison) repeats each line using the same rhythmic/melodic motif.
  • In canonic imitation, performer 1 starts with one phrase while performer 2 repeats the same words and melody. The performers thus repeat each other with a time delay.


  • The last syllable of a prosodic phrase is generally lengthened, and often at a low pitch.
  • There is one example of the lengthening of the last syllable of an interrogative phrase, and at a high pitch that might reflect speech intonation.

Word variations

  • Vocables occur.
  • Schwa vowels (normally u) have the same duration as all other vowels.

The two Seediq performances demonstrate many characteristics similar to other vocal expressions that are studied here. They can be understood as performance templates that give performers the possibility to vary the words or to add words, or to improvise new words ex tempore. In this case, the metre of the words changes from one stanza to another. It seems as though each new stanza or section is made up of a new phrase and that new lines are made up in the same metre for the remainder of that stanza, until a new phrase of a different metre commences. The realization of these lines of different metres is basically of two kinds: (1) a short motif in a rather fast pattern dominated by short tones in an iambic beat and ending with a long tone, and (2) slightly longer and slower phrases, including a long tone on the third syllable. These motifs and corresponding words are repeated in pairs. These pairs may be performed by two performers interlocked in a canonic imitation, so that the second performer starts later than the first, and this delay will last until the end of the performance.

1 Loh 1982, Tan 2012.
2 Internet reference: Seediq Bale song. For more information and controversy regarding historical representation, see
3 Loh 1982: 315–316. The Atayal and the Seediq are closely related both linguistically and culturally.
4 Loh 1982: 249–50.
  • Collapse
  • Expand

All of MUP's digital content including Open Access books and journals is now available on manchesterhive.


In the borderland between song and speech

Vocal expressions in oral cultures


All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 193 69 5
PDF Downloads 59 34 2