Printer Friendly

The poetics of central Australian song.

Abstract: An often cited feature of traditional songs from Central Australia (CA songs) is the obfuscation of meaning. This arises partly from the difficulties of translation and partly from the difficulties in identifying words in song. The latter is the subject of this paper, where I argue it is a by-product of adhering to the requirements of a highly structured art form. Drawing upon a set of songs from the Arandic language group, I describe the CA song as having three independent obligatory components (text, rhythm and melody) and specify how text is set to rhythm within a rhythmic and a phonological constraint. I show how syllable counting, for the purposes of text setting, reflects a feature of the Arandic sound system. The resultant rhythmic text is then set to melody while adhering to a pattern of text alliteration.


To understand how Australian Aboriginal performers combine text, music, dance and art, is, as Marett has stated, 'to approach the core of the poetics and aesthetics of their tradition' (2005:81). In Central Australia many songs are part of larger performance genres involving dance, gesture and visual designs that are associated with a particular tract of land, often referred to as a 'country' (Moyle 1986:3). (1) While dance, visual designs, hand-clapping, and the telling of associated narratives are not obligatory when performing such land-based songs, rhythm, melody and text are. By exploring how the essential components of song are woven together I hope to assist listeners to gain a deeper level of appreciation of the aesthetics of CA song. (2)

The approach here draws upon the ideas of Dell and Halle (in press), set out in their analysis of text setting in French and English songs, to show how a CA song is a composite of three independent structures: text, rhythm and melody. An important aspect of the composite is what is often referred to as constituency matching--the matching of units within each component (Dell and Halle in press). (3) In this article I identify the structure of these three independent components in Akwelye songs (pronounced akOOlya) and show that constituency matching between text and rhythm occurs within two constraints; the first affecting text and the second affecting rhythm:

* syllables must begin with a consonant

* rhythmic units (rhythmic cells) with fewer notes must not precede ones with more.

The rhythmic text is repeated until the melody is complete. In this process, the beginning and end of the composite rhythmic text is concealed. Ellis (1985:109-11) and Barwick (2000:331) have described such cyclic songs as having a 'timeless' aesthetic quality. (4) Timelessness is also a feature of the altyerre ('Dreaming')--the law that underpins Aboriginal society and all that that has created it, existing simultaneously in the past, present and future - which Stanner translates as the 'everywhen' (1953:24). (5)

Song as a composite

In most English songs, text is an independent structure to the 'tune'; tune having both rhythm and melody. Different verses (texts) can be put to the same tune and the song is regarded as the one song. In contrast, rhythm and melody are independent in CA songs. (6) Here substitution of a different text while maintaining the same melody and rhythm would be regarded as either incorrect or a different song. The different relationship between text, rhythm and melody in these two song styles is shown in Figure 1.



Central Australian songs often set the same rhythmic text to different parts of the melody. In a Central Australian song, text, rhythm and melody interact in a set order with constraints on how text is set to rhythm, and how the resulting rhythmic text is set to melody. Figure 2 shows these components of the song composite. (7)

Figure 2 (i) represents the process of aligning rhythmic and text groups, which I call 'text setting' or 'versification'. This produces a composite I call the 'rhythmic text'. It interacts with melody in a way that highlights its underlying structure (discussed below). Figure 2 (ii) represents the further process of aligning the resultant rhythmic text to the melodic groups, a process I call 'cantillation' (Turpin 2005).

One similarity between each of these three components is that they are all hierarchically structured. Syllables are grouped to make words and words are grouped to make text; and, as will be shown below, notes are grouped into rhythmic cells, which are in turn grouped to make rhythmic lines. Melody too is hierarchically structured.

An important aspect of the composite is what is often referred to as constituency matching -the matching of units within each component (Dell and Halle, in press). There is a strict match between text units and rhythmic units; and the resulting rhythmic text units match the divisions in the melody.


The Akwelye genre

Akwelye is a song series of the public women's ceremonial genre awelye. Awelye ceremonies are widespread throughout the Arandic-speaking area of Central Australia. (8) The Akwelye songs are owned by the people of Arnerre, a Kaytetye-speaking land-holding group of Central Australia whose main Dreaming is akwelye, meaning 'rain/rain cloud'. Kaytetye is one of the languages and dialects of the Arandic subgroup (9) of the Pama-Nyungan language family (Figure 3).


Akwelye songs are created by the Dreaming ancestors and revealed to humans in dreams. The songs are passed on from generation to generation, performed by a group of women led by the most senior woman. While the songs are undeniably owned and performed by Kaytetye-speaking persons, not all the words in the songs are from everyday spoken Kaytetye. (10)

Performance of an Akwelye song item lasts between 30 and 40 seconds, or longer than a minute with dancing (Figures 4 and 5). As is common throughout Central Australia, all songs within a song series have the same melodic structure. (11) A performance of a song series may involve singing over 80 songs, with a short interval of chatter and preparation between songs. (12)

Text structure

Like most Central Australian songs, the text of Akwelye songs repeats until the end of the melodic structure, at which point the song ends. (13) The song text is thus the smallest repeating unit of the total text of a song. An example Akwelye song text is shown in Figure 6, where it is represented as a hierarchy of nested groups, or constituents. Dell and Halle (in press) regarded such grouping structures in song as equivalent to prosodic units in speech.

Figure 6 shows that a complete song text (abbreviated as ST at the right edge of the top line) is a group (E), which is made up of two groups (C and D). (14) Following Barwick (1989:18), C and D are each called a text-line pair (TLP), as each is made up of two near-identical text-lines (TL). That is,

C is made up of two near identical text-lines [A.sup.1] and [A.sup.2]; (15) and D is made up of two near identical text-lines [B.sup.1] and [B.sup.2]. Within each text-line are smaller groupings of lexical words. Each A textline is made up of two words, N (kwerrepe-kwerrele) (16) and P (arenhepe-nheme); (17) while each B text-line is made up of three words, Q (pereltye), (18) R (ingketye) (19) and S (arrerne). (20) The next level of constituents shows phonological words (PW). For N and P this means a further division: N is made up of two phonological words T (kwerrepe) and U (kwerrele), and P is made up of two phonological words V (arenhepe) and W (nheme). Q, R and S are already phonological words, so the same unit as the lexical word is shown at this level.

The text-line is the largest group of non-repeating constituents. In contrast, song texts and text-line pairs are by definition made up of repeating text-lines. As will be shown the text-line is the significant unit when it comes to setting text to rhythm. In contrast, the text-line pair and the song text are the significant units when it comes to setting the rhythmic text to melody.


Elision of final vowels before a vowel-initial word

When spoken in isolation, all words in Arandic languages can be pronounced ending in 'e'. In flowing speech, however, the final vowel of a word is not usually heard when the next word begins with a vowel within the prosodic unit; for example, pereltye ingketye sounds perelty ingketye. As well, certain vowels are not pronounced in certain contexts. For example, the first 'e' in a sequence pere is often omitted, so a word such as pereltye is more often said preltye. The division of a song text into number of syllables is based on spoken pronunciations of the phonological phrase rather than individual words. For example, the number of syllables in pereltye ingketye is four (pre Ityi ngke tye) and not six (pe re Itye ing ke tye). (21)

The words making up the rest of the text-lines in Figure 6 are shown in Figure 7.

In Figure 7 the sung text is separated into syllables by spaces. In the row below, speech words are separated by a space and morphemes by a hyphen.

The constituents labelled W and S are the last words of the text-line and so the following vowel-initial word is outside of their prosodic unit. This means that W (-nhe-me) and S (arrerne) are not subject to the final-vowel elision rule (although their final vowels are subject to a pattern of vowel alliteration). In contrast, constituents U, Q and R and the following vowel-initial word all occur within the prosodic unit of the text-line. This means that the final e of the phonological words U (kwerrele), Q (pereltye) and R (ingketye) is replaced with the initial vowel of the following word ('a', 'i' and 'a' respectively). In song, where the end of a word often coincides with a long note, the initial vowel is audibly separated from the rest of the word. In Arandic languages the absence of an initial vowel is sometimes a different word altogether; for example, kwerre 'girl' and akwerre 'coolamon' in Kayetye. The shifting of initial vowels into the previous text unit complicates the identification of speech equivalents in song.



Song text structure

In Figure 6 we saw that a song text E is made up of two text-line pairs, which in turn are made up of two near-identical text-lines. The text-line structure for this song can be represented as AABB, which is referred to as a 'doubled' structure (Barwick 1989:18). (22) The naming of text-lines as A and B is not meant to imply that A must precede B. (23) Indeed, for most Akwelye song texts, it is possible to commence with either text-line pair (or either text-line in a AB structure). (24) While the vast majority of Akwelye song texts have this doubled structure, there are two other less common structures: AB and AAB. (25) Each song text is associated with only one of the three repetition configurations. So far I have identified the hierarchical structure of the song text as a sequence of phonological words which are grouped to form two text-lines. These text-lines are then associated with one of three patterns of text-line repetition (song text structure): AABB, ABB or AB

Rhythmic structure

Each Akwelye song has a set rhythm that is repeated until the end of the melodic structure. As with song text, song rhythm is the smallest repeating rhythmic unit in a song. The hierarchical structure of an Akwelye song rhythm is represented in Figure 9. This is the rhythm that accompanies the song text shown in Figure 6.


In Figure 9 the last row of 'x' marks the regular thigh-clapping beating that accompanies singing. Above this, the rhythm of the song is written in standard musical notation. The next row shows the smallest rhythmic units that recur throughout Akwelye songs. These are called rhythmic cells (RC) following Ellis (1985:93). (26) In Figure 9 the first three rhythmic cells are exactly the same and so are labelled the same: III. (27) In Akwelye, rhythmic cells are defined not only as the smallest recurring rhythmic unit, but are also the length of two crotchet beats.

The smallest repeating pattern of rhythmic cells is a rhythmic line (RL). (28) In Figure 9 the pattern of rhythmic cells III, III, III, II forms the rhythmic line VI, which is then repeated; and the pattern of rhythmic cells I, II, II form the rhythmic line VII, which is then repeated. Significantly, it is the rhythmic line (VI and VII) that is the largest non-repeated rhythmic grouping in a song.

Rhythmic lines repeat to form what I call rhythmic line pairs (RLP). These are labelled VIII and IX in Figure 9. Rhythmic line pairs also repeat to form the song rhythm (SR)--the smallest repeating rhythmic unit of a song. This is labelled X in Figure 9.


If the rhythmic line is made up of four or more rhythmic cells, Barwick (1989:18) recognised a further division of rhythmic lines into rhythmic segments. Division of the song rhythm in Figure 9 into rhythmic segments (RS) is shown in Figure 10, where it can be seen that only the first rhythmic line VI has rhythmic segments (RS): IV and V.

Figure 10 shows that rhythmic segments (IV, V) and rhythmic lines (VI and VII) are not made up of a repeated pattern, yet the larger groupings --rhythmic line pairs (VIII, IX) and the rhythmic song (X)--are: for example, VIII is made up of a repeated VI. Thus, rhythmic segments and rhythmic lines are the largest non-repeated rhythmic grouping of a song.

Matching text to rhythm

Akwelye has strict constituency matching between text and rhythm. Figure 11 shows the correspondence of text with rhythm of song text 42.


In Figure 11 the edges of the text units within E coincide with the edges of the rhythmic units within X. We can see that the number of constituents within all hierarchical levels of grouping is congruent between text and rhythm. In Figure 11 two phonological words and two rhythmic cells make up the rhythmic word segments (N, P, IV, V); two words and two rhythmic segments make up the A rhythmic text-lines ([A.sup.1], [A.sup.2], VI, VI) and three phonological words and three rhythmic cells make up the B rhythmic text-lines ([B.sup.1], [B.sup.2], VII, VII); two nearly identical text-lines and two identical rhythmic lines make up a rhythmic text-line pair (C, D, VIII, IX) and two text-line pairs and two rhythmic line pairs make up a rhythmic song text (E, X).

This strict constituent matching can be stated as: the number of (phonological) words in a song matches the number of rhythmic cells. In total, the rhythmic song text has fourteen phonological words and fourteen rhythmic cells, and the higher groupings of these constituents coincide throughout. Furthermore, not only is there strict constituency matching between text and rhythm, but where there is a pattern of repetition in the text then this is matched by a pattern of repetition in the rhythm (but not vice versa). This is seen at the level of the rhythmic text-line pair, where the matching involves a repeated text unit and a repeated rhythmic unit: [A.sup.1], [A.sup.2] is matched with VI, VI; and [B.sup.1], [B.sup.2] is matched with VII, VII.


In Akwelye a single syllable is not usually set to rhythmic notes. Constituency matching at the level of the syllable can be stated as: the number of syllables in a speech word matches the number of notes in the rhythmic cell. At first glance it may appear that this matching rule is not always upheld: for example, both pereltye and arrerne have three syllables, yet Figure 11 shows that they are both set to a two-note rhythmic unit. As I will show, syllable matching is complicated by two constraints that operate in Akwelye: a phonotactic constraint and a rhythmic ordering constraint. I suggest that the resulting permutations of words when set to a rhythmic unit can be understood in terms of the definition of a syllable in Arandic languages.

Phonotactic constraints in Akwelye

We saw that final vowels were deleted before a vowel-initial word within the prosodic unit due to a phonological rule in the language, and that in song this often results in a 'loss' of the initial vowel as it moves into the preceding rhythmic text unit. There is a phonotactic constraint in Akwelye that also affects the beginning of words and thus further conceals the speech equivalent. The phonotactic constraint occurs in the setting of text to rhythm and can be stated as: all syllables must begin with a consonant in Akwelye.

Arandic languages are somewhat unusual for Australian languages in that most words begin with a vowel rather than a consonant. The tension between this fact of the language and the syllable constraint stated above are resolved in Akwelye in one of two ways.

As discussed in the second section, when a vowel-initial word occurs within the text-line, the initial vowel becomes part of the preceding rhythmic text unit, displacing the final vowel. As a result the following consonant then begins the next rhythmic text unit. This was illustrated with song text 42, where we saw that the initial vowels of the speech words arenhepe-nheme, ingketye and arrerne are part of their preceding syllables, which are part of the preceding rhythmic text cells (UIII, QI and RII respectively). Thus, while the number of text units and rhythmic units is congruent within the consonant-initial syllable constraint, the edges of these units are not aligned when the word is vowel-initial. This is shown in Figure 12. (29)

In Figure 12 the right edges of text unit U and rhythmic unit III are not aligned; neither are they for units Q and I; nor R and II. As a result, the left edges of the following units are also not aligned: V and III, R and II, S and II. Such misalignment makes it difficult to identify vowel-initial speech words, especially words that have misalignment at both right and left edges, such as ingketye.

For vowel-initial words that begin a text-line there are two ways to avoid beginning with a vowel. One is to delete the initial vowel of the word and the other is to insert a consonant before the word. These two strategies are seen in textline A and text-line B of song text 37, shown in Figure 13. Text-line A is an example of vowel deletion, and text-line B is an example of consonant insertion.

In Figure 13, text-line A begins with a consonant by deleting the initial vowel ('a'), so that the rhythmic text cell is rntepe (CII), while text-line B begins with a consonant by inserting 'w' before the vowel-initial word aleme (EIII). Thus, while the number of text units and rhythmic units is congruent - two in text-line A and three in textline B--the consonant--initial syllable constraint means that the edges of the text and rhythmic units are not aligned when the word is vowel-initial. In song text 37, this affects all words, because all are vowel-initial, meaning that there is misalignment at both right and left edges of nearly all words. As a result of the phonotactic constraint, it is difficult to identify words such as aleme and itenyewe.





In Figure 13 it can be seen that deleting the initial vowel of a word creates one less rhythmic syllable in the unit than inserting a consonant. For example, the three-syllable word arntepe has the initial vowel deleted, creating only two rhythmic syllables (rntepe, CII). In contrast, the word aleme, versified by inserting a consonant, creates three rhythmic syllables (walemay EIII), yet had the initial vowel been deleted it would have had only two rhythmic syllables (lemay). The reason for choosing to insert a consonant rather than delete the vowel in this case is explained below.

Vowel-initial words have an ambiguous status as to their number of syllables when it comes to the matching rule, which states that the number of syllables in a word must equal the number of notes in the rhythmic cell. That is, the initial vowel may or may not be counted as a syllable when it comes to setting text to rhythm. The optional pronunciation of word-initial vowels is well attested in Arandic languages, and has even led to debates as to whether the syllable in Arandic languages is consonant-initial or vowel-initial (Breen 1990, 2001:49; Breen and Pensalfini 1999). For example, pereltye 'River Red Gum lerp' (attested in all Arandic varieties) can also be pronounced as apereltye. I suggest that the dual treatment of initial vowels in song, where it is sometimes counted as a syllable and sometimes not, reflects its optional status in spoken Arandic languages. That is, it is precisely because this is an area of flexibility in the spoken language that the poetic genre allows flexibility in how to count syllables in relation to vowel-initial words.

Hierarchy of notes/syllables in rhythmic text-lines

The significant factor as whether to count an initial vowel as a syllable or not seems to be the constraint affecting rhythmic lines already stated in the first section and repeated here: rhythmic cells with fewer notes can not precede ones with more. This can be seen when considering the total number of different rhythmic lines in Akwelye songs, as in Table 1.

Table 1 shows that there are 10 different rhythmic lines, labelled Z to Q in the left-hand column. The right-hand column shows the total number of different text-lines set to each rhythmic line. There are no instances of rhythmic cells with fewer notes preceding one with more notes, although a text-line may be made up of rhythmic cells with the same number of notes, as in Z, X, W and S. In terms of whether to count an initial vowel of a word as a syllable or not, and hence to set it to a two, three or four-note rhythmic cell, it is necessary to consider how many syllables are in the other words of the text-line. For example, in text-line 37B, the word aleme (three syllables) is followed by itenyewe (four syllables). Because a rhythmic cell with fewer notes must not precede one with more, it is necessary to count the initial vowel as a syllable for aleme, but not for itenyewe. This then gives an acceptable sequence of rhythmic cells: III plus III.

As well as these strict matching rules and rhythmic hierarchy, there is also an optional phonological rule affecting rhythmic text-lines, stated as: a syllable beginning with w, r or b occurring on the second note of rhythmic cell III may be elided.

For example, pereltye ([??] [crochet]) becomes preltye ([crochet] [crochet]). This optional rule is not a violation of the rhythmic hierarchy as it occurs after the formation of the rhythmic text composite. That is, the text must already be set to rhythmic cell III for this to occur. (30) Furthermore, the optional phonological rule has its counterpart in spoken language, as discussed in the second section.

In summary, the number of rhythmic and textual constituents at all structural levels of Akwelye must match. Repetition of text must also be matched with repetition of rhythm. Flexibility in how to count syllables for the purposes of text setting relates to a phonological feature of the spoken language; and the precise count for each text-line is motivated by a requirement that rhythmic cells with fewer notes must not precede ones with more.

The Akwelye requirement that all syllables must begin with a consonant (causing the addition or deletion of sounds at the beginning of a line and the misalignment of rhythmic and textual units within the line), in combination with the optional deletion rule stated above, means that only those well versed in the Akwelye rules of text setting can accurately identify the words in a song.

Melodic structure

The Akwelye melodic contour is made up of melodic sections which are defined by where a breath is taken (hereafter MS). (31) Figure 14 is a transcription of the first song of track 26 from the CD Awelye Akwelye: Kaytetye women's traditional songs from Arnerre. Breath intakes (,) and melodic sections (MS1, MS2, and so forth) are marked on the transcription in Figure 14.

Other than the introductory melodic section (which has its own melodic pitch), it can be seen that all subsequent melodic sections have the same pitch set as each other, and so melodic sections can also be identified by consistency in form. (32)

The characteristics of the five melodic sections are illustrated in Figure 15, following the model of Ellis (1998:435). The third and fourth melodic sections may be repeated to create a longer song. This is the melodic structure of all Akwelye song items; yet because rhythmic patterns vary within the one song series, there is variation in the number of notes sung to each melodic section across different song texts.

The Introductory MS is sung at an octave above the lower tonic (8ve]. The Descent MS begins a slightly flattened fourth above the lower tonic and descends to the tonic. The Internal MS begins with the same descent but has a much longer tonic. The Unaccompanied and Final melodic sections have the same melodic structure as the Internal melodic section, however the Unaccompanied MS has no accompanying thigh clapping.

Matching rhythmic text to melody

Barwick shows that the combining of a rhythmic text with melody in a Pitjantjatjara song adheres to particular structural principles of the text/ rhythm (1989:13). (34) Similarly in Akwelye, the structural boundaries of the rhythmic text must match the boundaries of the melodic units. While most frequently it is the higher level boundaries of the rhythmic text (song text, rhythmic text pair) that match the melodic groups, it is also possible to match the rhythmic text-line and rhythmic text-segment of a long text-line to a melodic group.

The melodic structure of the song in Figure 14 is shown in Figure 16.

In Figure 16, the labels MS1, MS2 (and so forth) refer to the melodic sections marked in Figure 14; the letters A and B refer to the rhythmic text (RT) lines and underlining shows that the melodic section is sung without accompanying clapping. It can be seen that the song consists of four repetitions of the song text: one beginning at MS1, the second at MS3, the third at MS4, and the fourth at MS5. The song ends with a gradual dropping out of voices during text-line A.

Figure 16 shows one way in which the rhythmic text is set to melody. The second song of track 26 begins with the B text-line, and so the setting of rhythmic text to melody is represented as in Figure 17.

Figure 18 shows a further setting of song text 33 from a different performance, where it can be seen that the end of MS3 aligns with the end of the RT line pair (AA), instead of the end of the rhythmic song text.

For RT lines that have six crotchet beats or more (rhythmic lines V-P in Table 2), the end of a melodic section can be aligned with a RT line rather than the RT line pair. (35) This can be seen in the various melodic settings of the rhythmic text of song text 7, track 3 of Awelye Akwelye, shown in Figure 19.

Figure 19 shows that in song item (i), each melodic section lasts for a RT line pair. In (ii) this holds for MS1 and MS2, but MS3 and MS4 have an additional RT line: AAB and BAA respectively. In (iii), MS3 is even longer, comprising a complete rhythmic song text (AABB). A large study of rhythmic text to melody setting in Akwelye shows that MS1 and MS2 require a match of either a RT line or RT line pair. In contrast, MS3 can be matched with one or more RT line pairs, which can be followed by a RT line or an even smaller unit, the RT line segment. In contrast, MS 4 requires a strict match of a RT line pair, which can be followed by a RT line. (36)

The process of setting an Akwelye rhythmic text to melody--cantillation--is not a single fixed composite, but a creative act where singers draw upon their knowledge of the structural principles of the rhythmic text and the melody, as other researchers have shown. (37)


Text setting in Akwelye requires constituency matching between words and rhythm within certain phonological and rhythmic constraints that can force alteration to the words. This marks the rhythmic text clearly as a non-speech genre. The performance of the rhythmic text to the melodic sections, with the patterns of alliteration realised in this process, is a further stage at which alteration to the original words occurs. It is no small feat to maintain the integrity of each component while at the same time interlocking them by matching the constituent groups.

The interlocking of rhythmic text to melody, while maintaining alliteration, draws attention to the form of the song in two ways: it identifies the song as 'fitting' the Akwelye genre; and in doing so, any deviations from these sound patterns are a 'signal' that a particular song has something unique. (38) Such adherence to and deviation from standard sound patterning carries with it an intangible aesthetic quality, which Jakobson calls the poetic function of verbal art (1987:71), (39) yet it comes at the expense of being able to convey semantic meaning, because the words are necessarily altered to fit the metrical genre.


I am grateful to many Kaytetye people who shared their songs and explanations of songs with me throughout this research, in particular, Daisy Kemarre, Katie Ampetyane, Blanche Ross, Amy Ngamperle, Lena Ngamperle and Hilda Ngamperle. I am indebted to Alison Ross for her assistance with linguistic transcription and translation. I am also grateful to Grace Koch who made available her field recordings for this research. An Australian Post-doctoral Award, a Sydney University fieldwork grant, and a post-doctoral fellowship by the Endangered Languages Documentation Program provided funding for this work.


Barwick, Linda 1989 'Creative (ir)regularities: The intermeshing of text and melody in performance of central Australian song', Australian Aboriginal Studies 1989/1:12-28.

-- 2000 'Song as an Indigenous art' in M Neale and S Kleinert (eds) Oxford Companion to Aboriginal Art and Culture, Oxford University Press, Melbourne, pp.328-34.

-- 2005 'Performance, aesthetics, experience: Thoughts on Yawulyu Mungamunga songs', in E Mackinlay, S Owens and D Collins (eds) Aesthetics and Experience in Musical Performance, Cambridge Scholars Press, Newcastle, pp.1-18.

Barwick, Linda and Allan Marett 2003 'Aboriginal traditions' in A Scott-Maxwell and J Whiteoak (eds) Currency Companion to Music and Dance in Australia. Currency House, Sydney, pp.26-7.

Breen, Gavan 2001 'The wonders of Arandic phonology', in S Simpson, D Nash, M Laughren, P Austin and B Alpher (eds) Forty Years On: Ken Hale and Australian Languages, Pacific Linguistics, Research School of Asian and Pacific Studies, Australian National University, Canberra pp.45-69.

Breen and Pensalfini 1999 'Arrernte: A language with no syllable onsets', Linguistic Inquiry 30:1-25.

Dell, Francois and John Halle (in press) 'Comparing musical text setting in French and English songs', in Jean-Louis Aroui and Andy Arteo (eds) Towards a Typology of Poetic Forms, Elsevier, Amsterdam.

Dussart, Francoise 2000 The Politics of Ritual in an Aboriginal Settlement: Kinship, gender, and the currency of knowledge, Smithsonian Institution Press, Washington DC.

Ellis, Catherine 1985 Aboriginal Music: Education for Living: Cross-cultural experiences from South Australia, University of Queensland Press, St Lucia.

-- 1997 Understanding the Profound Structural Knowledge of Central Australian Performers from the Perspective of TGH Strehlow, Strehlow Research Centre, Alice Springs (Occasional Papers 1) pp.57-78.

-- 1998 'Central Australia', in Adrienne L Kaeppler and JW Love (eds) Garland Encyclopaedia of World Music: Australia and the Pacific Islands, Garland Publishing, New York, pp.432-8.

Ellis, Catherine and Linda Barwick 1987 'Musical syntax and the problem of meaning in a Central Australian songline', Musicology Australia 10: 41-57.

Fabb, Nigel 1997 Linguistics and Literature, Blackwell Publishers, Massachusetts.

Glowczewski, Barbara 1999 'Dynamic cosmologies and Aboriginal heritage', Anthropology Today 15:3-9.

Hale, Kenneth L 1984 'Remarks on creativity in Aboriginal verse', in JC Kassler and J Stubington (eds) Problems and Solutions: Occasional essays in musicology presented to Alice M. Moyle, Hale & Iremonger, Sydney, pp.254-62.

Jakobson, Roman 1987 'Linguistics and Poetics', in K Pomorska and S Rudy (eds) Language in Literature, Harvard University Press, pp.62-94.

Keogh, Ray 1990 'Nurlu Songs of the West Kimberleys', unpublished doctoral dissertation, University of Sydney.

-- 1995 'Process models for the analysis of Nurlu songs from the western Kimberleys', in L Barwick, A Marett and G Tunstill (eds) The Essence of Singing and the Substance of Song: Recent responses to the aboriginal performing arts and other essays in honour of Catherine Ellis, Oceania Publications, University of Sydney (Oceania Monograph 46).

Koch, Grace 1987 'Dyirbal Gama songs of Cape York', in M Clunies Ross, T Donaldson and S Wild (eds) Songs of Aboriginal Australia, Oceania Publications, University of Sydney (Oceania Monograph 32), pp.43-62.

Koch, Harold 2001 'Basic vocabulary of the Arandic languages: From classification to reconstruction', in J Simpson, D Nash, M Laughren, P Austin and B Alpher (eds) Forty Years On: Ken Hale and Australian languages, Pacific Linguistics, Research School of Asian and Pacific Studies, Australian National University, Canberra, pp.71-87.

Marett, Allan 2005 Songs, Dreamings and Ghosts: The enactment of ancestral precedent in the wangga of north Australia, Wesleyan University Press, Middletown.

Moyle, Richard 1986 Alyawarra Music: Songs and society in a central Australian community, Australian Institute of Aboriginal Studies, Canberra.

-- 1995 'Singing from the heart?', in L Barwick, A Marett and G Tunstill (eds) The Essence of Singing and the Substance of Song: Recent responses to the Aboriginal performing arts and other essays in honour of Catherine Ellis, University of Sydney, pp.53-8.

Osborne, Charles Roland 1989 Tiwi chanted verse, unpublished manuscript, held at AIATSIS (call no. B O811.96/T2), Canberra.

Stanner, WEH 1953 'The Dreaming', in White Man Got No Dreaming: Essays 1938-1973, ANU Press, Canberra, pp.23-40.

Strehlow, TGH 1971 Songs of Central Australia, Angus & Robertson, Sydney.

Treloyn, Sally 2003 'Scotty Martin's Jadmi Junba: A song series from the Kimberley region of northwest Australia', Oceania 73:208-20.

-- 2006 'Songs that Pull: Jadmi Junba from the Kimberley region of northwest Australia', unpublished doctoral dissertation, University of Sydney.

Tunstill, Guy 1987 'Melody and rhythmic structure in Pitjantjatjara song', in M Clunies Ross, T Donaldson and S Wild (eds) Songs of Aboriginal Australia, Oceania Publications, University of Sydney (Oceania Monograph 32), pp.121-41.

Turpin, Myfany 2005 'Form and Meaning of Akwelye: A Kaytetye woman's song series of Central Australia', unpublished doctoral dissertation, University of Sydney.

Turpin, Myfany and Alison Ross 2004 Awelye Akwelye: Kaytetye women's songs from Arnerre, Central Australia (companion book to CD and cassette), Papulu Apparr-kari language and culture centre, Tennant Creek.


(1.) There are other types of traditional songs in Central Australia, many of which are not associated with countries, and some of which are sung individually. Such songs may be sung to alter weather conditions, people's behaviour, or recover from ill health. Other songs are sung purely for entertainment and others accompany narratives often involving sand-drawing. There are also songs sung by children, which are part of traditional games.

(2.) Strehlow, in his seminal work, Songs of Central Australia (1971), discusses the semantic complexities of many CA song texts. In contrast, I focus here on the relationship between text and rhythm.

(3.) Here I use 'groups', 'units' and 'constituents' as synonyms.

(4.) Ellis (1985:93), Moyle (1986, 1995), Ellis and Barwick (1987) and Barwick (1989) have provided discussions of the cyclic form of various Central Australian songs, as have Keogh (1995:41) and Treloyn (2003:213, 2006) for the Kimberley.

(5.) 'Dreaming' is a common English translation of the Aboriginal concept, rendered in Kaytetye as altyerre, which encompasses ancestral beings that manifest as plants, animals or natural features such as wind, fire, moon and the physical and social landscape that they created. The concept has been well described by Stanner (1953) and Glowczewski (1999:6).

(6.) The independence of rhythm and melody in Central Australian songs has been shown by Ellis (1985), Tunstill (1987) and Barwick (1989:14); and further west, for songs in the Kimberley region by Keogh (1990) and Treloyn (2003, 2006).

(7.) This ordering is suggested by the fact that the spoken version of the rhythmic text does not have the additional alterations to the text that exist when it is set to melody. Hale (1984), G. Koch (1987) and Osborne (1989) similarly recognised differences in spoken language and song texts in terms of processes.

(8.) Moyle (1986) has provided a discussion of Alyawarr awelye ceremonies. The cognate word in the neighbouring Warlpiri and Warumungu languages, yawulyu, refers to a similar genre of songs. Dussart (2000) and Barwick (2005) have provided descriptions.

(9.) This subgroup includes the currently spoken languages Arrernte, Alyawarr, Anmatyerr and Kaytetye, and various dialects of these languages. These, and other Arandic languages no longer spoken, have been discussed in Breen (2001). H. Koch (2001) has conducted comparative research on language/dialects within the Arandic family.

(10.) Many words are unattested in contemporary Kaytetye but exist in neighbouring Arandic languages. Others are unattested in any Arandic languages, and while some of these are said to be in an alternate register, such as a special song register, there are also some words that are simply unknown across the Arandic region.

(11.) It is perhaps for this reason that Strehlow (1971) referred to the song series as a 'song' and the individual songs as 'verses'.

(12.) The analysis is based on my research on a corpus of over 700 Akwelye songs, comprising 52 different song texts. These were performed over the course of 30 years, at five separate performance sessions. Grace Koch recorded performances of Akwelye in 1976 and 1994; the Central Australian Aboriginal Media Association (CAAMA) in 1986; and I recorded Akwelye in 1999 and again later that year with Linda Barwick. I recorded further performances of individual songs between 2000 and 2007.

(13.) The pervasiveness of this structure across central Australia is noted by Barwick and Marett (2003:26).

(14.) My labelling here follows the convention of using 'A' and 'B' for the text-lines of Aboriginal songs. Sequential labelling is then used for larger units (C, D and E) followed by smaller units (N to S).

(15.) The superscript signifies that the two text-lines are nearly identical, the only difference being a pattern of vowel alliteration.

(16.) kwerre 'girl' in Kaytetye, Western Anmatyerr and archaic in Arrernte. In Arrernte -pe- is an increment in linked reduplication signalling a continuous state.

(17.) are-nhe-penhe-nhe-me is glossed as follows: are-'see'; -nhe-pe-nhe 'do something while moving relative to some place or thing' -me 'present tense' in Arrernte.

(18.) (a)pereltye 'sweet lerp that grows on the leaves of the River Red Gum (Eucalyptus camaldulensis)' in all Arandic languages.

(19.) ingketye 'precious' in Anmatyerr and Arrernte.

(20.) The alteration of the final vowel of text-line final words (e.g. arrerne) is because this position is subject to alliteration--'Matching rhythmic text to melody' (above).

(21.) Song texts also involve various complex patterns of alliteration that affect the quality of vowels (Turpin 2005:220-45).

(22.) Here I am putting aside the minor differences between the repeated text-lines. I return to these differences in the next section.

(23.) The basis for calling one text-line 'A' and the other 'B' in the other two song text structures is complex and discussed in Turpin (2005:102).

(24.) For song texts whose lines are of different lengths, there is a tendency to commence with a particular text-line. However, the majority of song texts have lines of the same length and for these it is possible to commence with either text-line, a phenomenon known as reversibility (Turpin 2005:155ff.).

(25.) The repeated text-line in a three-part song text is always treated as the A text-line, hence there are no ABB song texts.

(26.) This is what Barwick (1989:17) called a 'beating cell'.

(27.) I use the labels I and II for the cells with two notes, III for the cell with three notes, and IV for the cell with four notes.

(28.) In the literature these are also referred to as isorhythmic lines; Ellis (1985) and Barwick (1989) have provided examples.

(29.) Hale (1984) identified a related phenomenon in the setting of Anmatyerr words in a Warlpiri song. He stated that the last consonant of a text-line becomes the first consonant of the repetition of that text-line (A1A2). Such a phenomenon may also be motivated by a requirement that songs begin with a consonant, because the Anmatyerr word beginning the text is vowel-initial.

(30.) Details of this optional phonological rule have been discussed by Turpin (2005:215ff).

(31.) The coincidence of breath points with structurally significant points in the melody is noted in other Central Australian songs--Moyle (1986:158); Tunstill (1987:128); Barwick (1989:17). In contrast, breath intakes do not always coincide with melodic sections in the Pitjantjatjara Ngintaka song series (Barwick 1989:16).

(32.) Note that the final syllable of a text-line is often shortened or even omitted before a breath intake.

(33.) I am grateful to Linda Barwick for setting out the transcription in Finale.

(34.) Strehlow (1971:40), Ellis (1985:105), Tunstill (1987:128), Keogh (1995) and Treloyn (2006) have discussed the complexities involved in matching a rhythmic text to melody in other traditional Australian songs.

(35.) This may be related to the need to take a breath.

(36.) Akwelye songs can have more than five melodic sections. In such cases there is a pattern of repeating the structure of MS3 and MS4, where MS3 shows great flexibility in matching RT, while MS4 shows less variation. Note that because songs end with a fade-out, it is difficult to hear the last RT line of a song.

(37.) Ellis (1997:61) has discussed a similar phenomenon in Pitjantjatjara songs.

(38.) Barwick (1989) and Treloyn (2006) also noted the aesthetics of irregularity as a feature of Aboriginal songs in Central Australia and the Kimberleys.

(39.) Also Fabb (1997:15, 145).

Myfany Turpin

Hans Rausing Postdoctoral Fellow, University of Queensland School of Social Policy and Research, Charles Darwin University, Alice Springs

Dr Myfany Turpin is a post-doctoral research fellow in linguistics at the University of Queensland where she is currently documenting songs of the Arandic language group of Central Australia. She received her doctorate from the University of Sydney for her thesis on the language and music of Kaytetye women's songs. She has also produced a Learner's Guide to Kaytetye (2000) --the most northern Arandic language--and dictionary (forthcoming), as well as articles on Kaytetye syntax, semantics and lexicography. Other research interests are documenting social and environmental indicators in Central Australia, where she is based at Charles Darwin University in School of Social Policy and Research, Alice Springs. She is also involved in language and music education in Aboriginal communities with the Centre of Australian Languages and Linguistics at Batchelor Institute of Indigenous Tertiary Education.

Table 1: Rhythmic lines in Akwelye

Rhythmic Rhythmic line
line Id Cells
 1st 2nd

Z [quaven][crochet] [quaven][chochet]
Y [??][crochet] [quaven][chochet]
X [??][crochet] [??][crochet]
W [quaven][chochet] [quaven][chochet]
V [??][crochet] [quaven][chochet]
U [??][??] [??][crochet]
T [??][crochet] [??][crochet]
S [??][crochet] [??][crochet]
R [??][??] [??][crochet]
Q [??][crochet] [??][crochet]

Total number of rhythmic lines

Rhythmic Rhythmic line
line Id Cells
 3rd 4th

W [quaven][chochet]
V [quaven][chochet]
T [quaven][chochet]
S [??][crochet]
R [quaven][chochet]
Q [??][crochet] [quaven][chochet]

Total number of rhythmic lines

Rhythmic Number Rhythmic cells Number of
line Id of notes making up line text-lines set to
 rhythmic line

Z 4 II + II 3
Y 5 III + II 38
X 6 III + III 28
W II + II + II 1
V 7 III + II + II 4
U IV + III 1
T 8 III + III + II 4
S 9 III + III + III 4
R IV + III + II 2
Q 11 III + III + III + II 5

Total number of rhythmic lines 90

Figure 16: The setting of rhythmic text (RT) to melody
of the first song of track 26 (song text 33)


Figure 17: The setting of rhythmic text to melody
of the second song of track 26 (song text 33)


Figure 18: The setting of rhythmic text to melody
of a further performance of song text 33


Figure 19: The three settings of rhythmic text to
melody in track 3 (song text 7)

(i) MS1 BB (ii) MS1 AA (iii) MS1 AA
COPYRIGHT 2007 Australian Institute of Aboriginal and Torres Strait Islander Studies
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2007 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Turpin, Myfany
Publication:Australian Aboriginal Studies
Geographic Code:8AUST
Date:Sep 22, 2007
Previous Article:Flesh with country: juxtaposition and minimal contrast in the construction and melodic treatment of jadmi song texts.
Next Article:Budutthun ratja wiyinymirri: formal flexibility in the Yolnu manikay tradition and the challenge of recording a complete repertoire (1).

Related Articles
Balgo: The Musical Life of a Desert Community.
A.L. Lloyd in Australia: some conclusions.
Wax Poetics: Hip-Hop, Jazz, Funk & Soul.
Burning interiors; David Shapiro's poetry and poetics.
Emblemata Sacra; the rhetoric and hermeneutics of illustrated sacred discourse.
The Environmental Imagination.
Musical and linguistic perspectives on Aboriginal song.
Australian Aboriginal song language: so many questions, so little to work with.
Jurtbirrk Love Songs from North Western Arnhem Land.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters