iDIDJ Australia Didgeridoo Cultural Hub

Iwaidja Jurtbirrk songs: bringing language and music together.

Article from:Australian Aboriginal Studies
Article date:September 22, 2007
Author: Barwick, Linda ; Birch, Bruce ; Evans, Nicholas

Abstract: Song brings language and music together. Great singers are at once musicians and wordsmiths, who toss rhythm, melody and word against one another in complex cross-play. In this paper we outline some initial findings that are emerging from our interdisciplinary study of the musical traditions of the Cobourg region of western Arnhem Land, a coastal area situated in the far north of the Australian continent 350 kilometres northeast of Darwin. We focus on a set of songs called Jurtbirrk, sung in Iwaidja, a highly endangered language, whose core speaker base is now located in the community of Minjilang on Croker Island. We bring to bear analytical methodologies from both musicology and linguistics to illuminate this hitherto undocumented genre of love songs.

Introduction

Jurtbirrk is the Iwaidja name for a genre of didjeridu-accompanied songs that are often referred to in English as 'love songs'. Jurtbirrk songs are individually composed and owned, usually inspired by particular recent events. Their texts are in the Iwaidja language, one of the highly endangered languages spoken in northwestern Arnhem Land. (1) The song corpus consists of thirty-two song texts recorded by the Iwaidja Documentation Project (2) team at Minjilang during 2003 and 2004. Since the beginning of the project in July 2003 we have recorded one hundred and thirty-six Jurtbirrk song items over the course of eight performances, representing thirty-two distinct song texts. Full details of these performances are available in the booklet accompanying the Jurtbirrk CD (Barwick et al. 2005). (3) The collaborative work between linguists and the musicologist in the Iwaidja Documentation Project has produced a large corpus of well-described and annotated song texts, allowing us to investigate in some depth questions regarding the interrelationship of musical and linguistic structures in the songs. After setting out general historical, ethnographic and social background on the Jurtbirrk song-set, its composers and its themes, we comment on the song corpus from the twin perspectives of musicology and linguistics.

The relationship between music and language, and between the academic disciplines that attempt to describe and explain their respective ontologies, is called into question in any analytical engagement with song. While we believe that there is no inherent incompatibility between the two methodologies, a mono-disciplinary approach to the complex object of study might consciously or unconsciously avoid tackling aspects of song that are perceived as falling under the aegis of the other discipline, and thereby run the risk of failing to take full account of the phenomenon. Each discipline is naturally aware of the benefits of the other: there have been and continue to be numerous instances of cross-disciplinary collaboration in both theoretical and applied contexts.

The application in musical analysis of methodologies from the fields of linguistics, psychology and information theory was of particular interest in the second half of the twentieth century, as can be seen in the work of musical analysts such as George P Springer (1956). Ian Bent and Anthony Pople (2001 particularly Section II.5 'History, 1945-70') discussed these theoretical developments from the perspective of musical analysis, while Howard and Datteri (2004) provided a useful recent summary of cognitive psychology literature on music and language. Musical hermeneutics, the analysis of musical characteristics in relation to social context, is fundamental to the sub-discipline of ethnomusicology but impossible without rich linguistic, ethnographic and other contextual information.

In linguistics, recent years have seen the emergence of language documentation projects that aim to record language use in as wide a range of contexts as possible (Himmelmann 1998; Woodbury 2003), and music is very often one of the genres most often suggested for documentation by community collaborators (Barwick 2006a). In addition to providing rich data for linguistic analysis of various types, language documentation projects like the Iwaidja project can also provide a sound basis for establishing a fine-grained understanding of other domains of knowledge within a given society, especially if specialists from these domains participate in the documentation project. Barwick's involvement in the Iwaidja research team, along with the musical expertise and interests of other members and community enthusiasm for the documentation of their songs, has led to the collection of a considerable quantity of song recordings and discussions about songs and their cultural significance, which we draw upon in this article.

There is a long history of collaboration between linguistics and musicology in the study of Australian song: from Strehlow (1971) and Ellis (1964), Hercus and Ellis (Ellis et al. 1966), A Moyle (1974) and Stokes (Stokes and Aboriginal advisers 1981) through to, in more recent times, Dixon and Koch (1996), Hercus and Koch (1999) and many others, including Marett, Barwick and Ford (2001). Scholars trained in both disciplines, such as Turpin and O'Keeffe (both with contributions to this volume) are also emerging. The work presented in this paper, therefore, takes place within a long and ongoing Australian tradition of cross-disciplinary engagement.

Diversity of language, diversity of song

Western Arnhem Land is famous for its linguistic diversity. The area bounded by Pine Creek in the south, the Mary River in the west, and the Liverpool and Mann Rivers in the east, and encompassing Cobourg Peninsula and the islands of Croker and Goulburn is home to around a third of the continent's twenty-five language families. (4) This is roughly the area of territory associated with a single medium-sized language such as Warumungu or Wajarri elsewhere in the continent. Figure 1 shows the approximate locations of the Iwaidjan languages in the northwestern part of Western Arnhem Land, on the Cobourg Peninsula and surrounding areas.

There are many competing, though not necessarily mutually exclusive, hypotheses that attempt to explain the extreme linguistic diversity of this area of the Australian continent. The size of social groups, time depth of settlement, isolation or interaction between groups, accretions of multidirectional borrowings, and the need for small groups to maintain distinct cultural and linguistic identities in order to assert their rights to Country, have all been put forward as contributing factors. Linguistic diversity clearly does not arise in a social vacuum. It is cultivated and maintained by sociolinguistic practices, and supported by cultural beliefs: for example, the beliefs that water sources become accessible only if addressed in the language associated with the country in which they occur, or that spirits assist those who speak the 'right' language, but harm those who do not (Brandl and Walsh 1982; Trigger 1987).

[FIGURE 1 OMITTED]

Staggering as the linguistic diversity is, however, the multiplicity of song styles in western Arnhem Land is even greater. Examining this rich musical tradition provides us with the opportunity to study cultural and linguistic diversity from a new perspective, as the elements of song are selected and curated more consciously than those in speech. An initial stocktake of the Cobourg area reveals that the number of named and musically distinct song-sets exceeds the number of identifiable languages (Table 1). A song-set is always associated with a particular language: Manbam with Marrku, Milyarryarr with Ilgar, Yanajanak with Amurdak, Ngarnarru with Manangkardi, Inyjalarrku with Mawng, and so on. Each language is associated with at least one distinct song-set, even languages that no longer have fluent speakers. This is the case, for example, for the Ilgar Milyarryarr songs and the Manangkardi Ulurrunbu, Mirrijpu and Ngarnarru songs.

The song-sets listed in Table 1 form part of the overarching musical corpus referred to in Iwaidja as manyardi--roughly translatable as 'songs'. There is a broad division between 'true' songs received by known individuals from beings resident in the country--these comprise the 'sea song' genre (known as ldalha 'sea' in Iwaidja, or kurrula 'sea' in Mawng), and the 'stone country song' genre (known as wardyad 'stone country' in Iwaidja)--and songs composed spontaneously (that is, without spirit intervention) by known individuals, often inspired by local events. The 'love song' genre--comprising the Iwaidja Jurtbirrk song-set (the focus of this article), the Itpi-itpi song-set (in Kunwinjku, Kun-barlang and Mawng languages) and at least one song-set in Kunwinjku, the Kun-nalk ('crying songs')--belongs in the latter category, along with church songs (Iwaidja Yiwarruj) and guitar band songs such as those composed in Kunwinjku and English by the Bininj Band, whose main performers are resident in Minjilang. As a general rule, spontaneously composed songs (that is, songs in the love song, church song and guitar band genres) are 'in language', composed of glossable grammatical statements in one or more locally spoken languages. By contrast, with few exceptions, 'true' songs (in both the sea song Idalha/kurrula and stone country wardyad genres) tend to be sung with untranslatable fixed syllable strings that are 'for song', although the songs often have a particular reference or story associated.

Iwaidja have shared in ceremonial and informal song performances with other groups from western Arnhem Land in various places and ways. The earliest recordings of Iwaidja singers were in Oenpelli (Kunbarlanja) in 1948 (Mountford 1949), Beswick in 1961 (West 1963), and in Bagot (in Darwin) in 1962 (Moyle 1967). In each instance these performances took place in camps where Iwaidja had gathered with members of several other different language groups, whose song traditions were also recorded. Each of these communities today continues to be multilingual, as do all the main communities in which Iwaidja is spoken, and there is a continuing practice of ceremonial exchange throughout the western Top End by which public ceremonial occasions such as funerals are marked by importing singers from another community. (6)

The western Arnhem Land region is of great musicological interest, having links to three contiguous but musically and ceremonially distinct areas: central and eastern Arnhem Land, the Daly region, and the Tiwi Islands, as is revealed in Alice Moyle's mapping of Australian musical styles (Moyle 1974). The music of western Arnhem Land has been relatively sparsely documented by musicologists in recent years, compared to the considerable attention paid to the musics of neighbouring regions (especially northeastern Arnhem Land and the Daly region). To our knowledge Iwaidja Jurtbirrk songs have never previously been recorded and published, although other Iwaidja songs in the Idalha 'sea song' genre were recorded by Simpson, West and Moyle (in the recordings previously mentioned). Given the multilingual social contexts in which Jurtbirrk songs have been created and performed, it is perhaps not surprising that they are similar in style to, and share musical components with, other public didjeridu-accompanied songs of the western Top End (for example, the kun-borrk of western and southern Arnhem Land, or the wangga and lirrga of the Daly region). Our discussion of Jurtbirrk songs therefore includes comparisons with recent work on wangga and lirrga by Marett and Barwick respectively (Barwick 2003, 2006b; Marett 2005) and on kun-borrk by Garde (2006, 2007).

The Jurtbirrk song-set

Jurtbirrk songs are sung in Iwaidja by one or two men (one of whom is often the composer), who accompany themselves on clap-sticks and are supported by another man playing didjeridu in the 'Western' style (using a rhythmically patterned drone probably involving vocalising pitches as the player breathes into the didjeridu, and without the overblown hoot characteristic of northeastern Arnhem Land styles--Jones 1963, 1967). On ceremonial or celebratory occasions, women and men may dance (in separate groups), but this is optional, and the songs may equally be performed in informal social contexts, without dancing.

As shown in Table 2, our sample includes thirty-two songs by four different living composers: David Minyimak, Reggie Cooper, Ronnie Wandijak and Robert Cunningham. Appendix 3 contains a listing of the songs in the corpus.

It is usual for the composer to lead singing of his own songs, but our corpus includes several song sessions that diverge from this convention. With twenty recorded songs, David Minyimak is the most prolific composer, but because his voice is no longer strong he usually asks another singer, Ronnie Wandijak or Reggie Cooper, to lead singing of his songs even when he is present, and both Wandijak and Cooper have authority to perform Minyimak's songs even when he is not present. Cooper's renditions of Minyimak's songs sometimes differ so greatly from the composer's own renditions that they can be considered as different compositions based on the same textual material. Thus, in addition to six recorded songs of his own composition, Cooper frequently performs his own adaptations of a further three songs originally composed by Minyimak. Ronnie Wandijak recorded his own four compositions interspersed with other songs originally composed by Minyimak and Cooper. The two songs composed by Robert Cunningham were both recorded for us by Reuben Arramunika, who also sang songs composed by Cooper and Minyimak but who did not record any songs of his own composition. Cunningham himself lives in a remote location at some distance from Minjilang and was unable to participate in our recording sessions or discuss his songs with us.

Jurtbirrk songs are spontaneous concoctions, along the lines of the blues songs that emerged from African-American culture in the twentieth century, and take as their subject intimate moments from everyday life. Composed by known individuals in the Iwaidja language, the songs typically deal with romantic or emotional topics. They are composed of brief, evocative statements in accessible language, usually consisting of two lines of text, which are repeated in predictable combinations. Jurtbirrk may be composed and performed by anybody, as long as they are male. There are no known instances of a woman composing or performing a Jurtbirrk song, although women very much enjoy listening to, and talking about, the songs and the events portrayed in them, and also compose dances for and dance to Jurtbirrk.

Jurtbirrk songs tell their story with great economy. The man who looks over his shoulder, waving and pointing ('Yangkuwilbarrjiny' JU13) (7) does that, and only that. The lyrics contain no further elaboration, and no names are mentioned. Table 3 shows a typical example, here presented showing full textual repetitions within the verse.

As explained to us by one of the senior composers, David Minyimak, Jurtbirrk songs are inspired by actual events, rather than received from spirits during sleep (this distinguishes Jurtbirrk from the ldalha sea-song repertories performed at Minjilang today, which are dream-composed). The events recounted in Jurtbirrk songs happened in an actual place and at a particular time. For those who have knowledge of the details, performance of the song brings back the memory of the circumstances of its composition. For those who are too young to know about it, the story may be told once again.

The exact dates for the composition of the individual Jurtbirrk songs we have recorded are difficult to pinpoint, but the composers date the oldest songs from what is referred to as 'Cape Don Time'. This is the period roughly from World War II through to the early 1960s when many Iwaidja speakers were based at the lighthouse settlement at Cape Don, at the western end of Cobourg Peninsula. With the gradual relocation of most of the Iwaidja-speaking population to Croker Island during the fifties and sixties, Minjilang became the most important location for the composition of Jurtbirrk, a situation that continues to the present day.

It is highly likely that Jurtbirrk and other 'love song' genres recorded by the Iwaidja project team in Minjilang are cognate with the 'gossip songs' recorded and transcribed by Catherine Berndt and Ronald Berndt in the Oenpelli and Goulburn Island areas in the 1940s (Berndt and Berndt 1951). During the Berndts' fieldwork in the Oenpelli and Goulburn Island areas from 1947 through to the 1980s, 'gossip' songs were common among Mawng, Kunwinjku and other associated language groups of the region. As Ronald Berndt (1987:172) described them, 'gossip songs' are:


... always about human beings and contemporary events; the messages
they convey are relatively straightforward, except that they do not
state who the participants in any one song really are.
Many characteristics of the Iwaidja Jurtbirrk song-set match this description. These public song-sets are to be distinguished from the various restricted song genres, sometimes also referred to in English as 'love songs,' that are or have been widespread across Aboriginal Australia (for example, jarrarta of the Roper region, or yilpinji of Central Australia) (Berndt 1978). Many of the latter are private or restricted in nature and not suitable for public performance.

Jurtbirrk songs are mostly about sexual love. They are peopled with characters like the young man combing his hair as he goes to meet 'someone' across the other side of the creek at Cape Don ('Nganayalkbarrki', JU29); or the woman who approaches a man as he walks along, grabs him by the chin, and kisses him ('Yakaldadbarjan', JU27); or the man who drifts off in mid-sentence because he's distracted by thoughts of his lover ('Wurruwarr', JU08). Jurtbirrk also takes as its subject matter the difficulties of relationships and liaisons. Jealousy is a recurrent theme. One composer sings about his women fighting among each other ('Ayunman wingalmu', JU18), another about being ignored by a woman who gets into another man's car ('Yarildariki', JU16). Still other songs tell of sulking ('Yangmanara' JU03), violence ('Ayiyakanjildiny', JU24), and the packing of bags and leaving home ('Riwujbakba', JU12).

More rarely, a Jurtbirrk song tells of something not obviously connected with love, like the image of a pearl lugger leaning over in a stiff breeze, from a song composed at Ldingi Point near Cape Don ('Dayibabu', JU21). A recent performance of this song brought back memories of the days when such 'diver boats' were a common sight around the bays and inlets of the Cobourg Peninsula, but nothing was mentioned that directly connected the song to a love affair or a sexual liaison (Table 4).

Musical analysis of Jurtbirrk Musical conventions

Songs in the Jurtbirrk corpus use four different rhythmic modes (10) defined by combinations of clap-stick tempo and vocal metre (Table 5). Two clap-stick tempo bands are used: 94 beats per minute and 105-115 beats per minute. The slower tempo band is associated with songs in simple metre (that is, with a consistent duple subdivision of the beat, transcribed in 3/4 or 4/4), and the faster tempo band with songs in compound metre (that is, those with an uneven or triple subdivision of the beat, usually transcribed in 9/8 or 12/8). (11) Songs within each of these broad categories can be further subdivided according to whether the text line occupies groupings of three or four beats (3/4 and 9/8 songs contrasting with 4/4 and 12/8 songs), yielding a four-category rhythmic modal system.

As can be seen from Table 5, David Minyimak's songs cover all four rhythmic modes in this system, while the other three composers use only one, namely 9/8 songs in 105-15 tempo band.

Typically for western Arnhem Land didjeridu-accompanied songs, the final note of all Jurtbirrk songs matches the didjeridu fundamental (an octave lower than the final of the voice). There is some variation in the scales used. (12) Songs performed by Reggie Cooper and Robert Cunningham are mostly sung in the Aeolian or natural minor melodic mode (a descending series built on the didjeridu fundamental E would be E', D, C, B, A, G, F#, E), and Ronnie Wandijak's are mostly in Phrygian mode (which differs from the Aeolian by using a minor second above the tonic: so the corresponding Phrygian descending series would be E', D, C, B, A, G, F, E). David Minyimak's songs sometimes use modes with a major third above the tonic (Mixolydian, Lydian and Ionian modes).

Great variety can also be seen in David Minyimak's use of textual repetition patterns and melodic contour, while the other three composers tend to use a single melodic mode, textual repetition pattern, and melodic contour throughout their repertory. One characteristic unique to the two songs composed by Robert Cunningham is the use of more than one clap-stick pattern within the same song: for the third line of text, these two songs employ a gapped pattern, in which the clap-sticks beat at approximately half speed (Figure 2, Musical notation of Robert Cunningham's song 'Kartbirljuju').

As previously mentioned, when Reggie Cooper performs songs by David Minyimak, he some times alters textual repetition pattern, melodic mode and melodic contour to conform to the conventions of his own repertory, to the extent that even those familiar with the repertory identify these songs as being new compositions by Reggie himself.

The strong correlation of composer with a particular cluster of musical characteristics is consistent with results of Marett's work on wangga songs in the Daly region (Marett 2005:194-5 and elsewhere). Marett (2005:132-3; this volume) also noted a reduction in musical complexity in recent performances of wangga compared to older recordings, which may suggest that the great diversity of musical features we have observed being deployed by Minyimak represents an older musical practice in the Jurtbirrk tradition. More study is planned on the correlation of musical features with particular composers and performers, and the question of whether such differences can in turn be related to the musical characteristics associated with particular genres, languages or country.

A diagrammatic representation of the structural components of Jurtbirrk performances, song items and verses is set out in Figure 3. Each song item (defined as a continuous stretch of musical performance) occurs in the context of a performance, and successive song items are interspersed with informal discussion. As a performance progresses, new songs are introduced, each usually repeated once or twice in successive song items before moving to the next song. There is no strict ordering of songs in the set, although sometimes successive songs may have a common theme or melody. Across the whole corpus individual songs recur in different performances: the most common song, 'Yangmanara' (JU03), is performed in 20 different song items.

As is common in public didjeridu-accompanied songs throughout Western Arnhem Land and Daly regions, each song item is strophically organised into verses. Didjeridu and clap-sticks play throughout the item, while the vocal part consists of two or three verses per item, each of which repeats more or less exactly the same song text. With some notable exceptions, the same song text is performed with identical musical setting across items and performances. The sample on which we draw consists of a total of 351 verses across the 136 song items; because of the repetition of verses within an item, we have at least two renditions of every song text in the corpus. In the case of the most commonly performed song, 'Yangmanara' (JU03), we have complete renditions of the song text in 56 verses across the 20 song items in our sample.

[FIGURE 2 OMITTED]

The third part of Figure 3, which shows the internal components of the Jurtbirrk verse, is exemplified with reference to 'Kurrana' (JU05) (other Jurtbirrk songs may vary in the number and constitution of components at the various levels). In this case, the verse is composed of two sections. Section 1 contains two presentations of the first line of text, the second truncated (as indicated diagrammatically by the section boundary ending before the line boundary in Figure 3--further discussion of this feature below). Section 2 contains line 2 followed by a string of vocables (marked 'voc' in Figure 3). Further detail on the internal constituents of 'Kurrana' (JU05) is set out in Section 3.3 below.

Musical expression of the rhetorical structures

Rhetorically, 'Kurrana' (JU05), like many other Jurtbirrk songs, is organised into an AAB pattern reminiscent of Afro-American blues songs, in which a repeated statement is capped by a second statement extending or explaining the first (Carroll 2005; Oliver 2001). As in the blues, the rhetorical structure of the Jurtbirrk song text is supported by its rhythmic and melodic setting, as can be seen in Figure 4.

[FIGURE 3 OMITTED]

Rhythmic setting: In this song each word in the text is co-extensive with one metrical unit of three clap-stick beats (here, transcribed as a 9/8 bar). In other songs that include words of one or two syllables two words may be presented over one metrical unit. The two-word text of line 1 (a)rdalbardalba yuwuldakbalkba is repeated to form section 1, but the second repetition omits both the unaccented first syllable a- and the final two syllables -balkba. The core syllables--rdalbardalba yuwuldak--are set over an identical rhythm each time. The unrepeated three-word text phrase of the second line is supplemented by a generic added vocable text ri ri ri. Similar vocables occur in this position at the end of every Jurtbirrk song verse.

Melodic setting: The verse is divided into two sections, divided by the breath taken at the end of the second rendition of the first text line. Each section has an overall descending melodic contour within the Aeolian (natural minor) melodic mode built on the didjeridu-defined tonic E: the first section oscillates from the fifth (B) to the third scale degree (G), with the last two metrical units performed entirely on G, while the second section covers the descent from the seventh (D) to the tonic E, with the last two bars performed entirely on the final E.

[FIGURE 4 OMITTED]

The unresolved feel of the first section's cadence on G mirrors the rhetorical structure, in which the unfinished repeat of the first statement leads strongly into its explication in the second line, where it becomes evident that the reason for the haste expressed in the first line is the imminent moonrise, which will reveal the lovers' movements to those watching. The resolution to the tonic E on the last word of the text--kurrana 'moon' --also mirrors the rhetorical structure, revealing at last the cause of the lover's anxiety. Marett (2005:1923) discussed similar ways in which the 'melody supports and reflects the structure of the text' in a wangga song.

Versification processes

We have classified as 'versification processes' additions or deletions of text syllables apparently motivated by musical factors. Partial reduplication: An example is:

(1) ngaldalmalangkaj [right arrow] ldalmalangka-yangka ('Nganbaldakaniny' JU09)

Partial reduplication like this occurs in only one song in the Jurtbirrk song repertory. By contrast, other song repertories described by linguists have a large number of reduplications--for example the Marri Ngarr Lirrga song texts analysed by Lysbeth Ford (2006).

Elision of syllables: A more regular feature of Jurtbirrk songs is the dropping of the final two syllables at the end of the first section. For example, in 'Kurrana' (JU05, cited in Figure 4 above), the repetition of the first line is truncated at the end of the second line, and the 'unsung' syllables are replaced by a vocal rest.

(2) ardalbardalba yuldakbalkba / rdalbardalba yulda' ('Kurrana' JU05)

Vocable insertion: Vocable insertion also occurs within some metrical units. The unit on which the vocables appear to be based is the syllable ya. This syllable is most commonly inserted once in a line, and is located at the end of a metrical unit, typically in instances where the preceding word, or combination of words, constitutes four syllables or less. The data also contain examples where the syllable ya appears to harmonise with the vowel in a preceding syllable, resulting in the variants yu and yi (examples (6) and (9) below).

(3) kudnuka [right arrow] kudnuka-ya ('Kudnuka ngartung' JU01)

(4) jumung mana [right arrow] jumung mana-ya ('Kartbirljuju' JU10)

(5) angkad birta [right arrow] angkad birta-ya ('Kudnuka ngartung' JU01)

(6) angkiju [right arrow] rdangkiju-yi ('Angkiju' JU20)

(7) yarukung [right arrow] yaruku-ya ('Angkiju' JU20)

(8) anayanjing [right arrow] rranayanji-ya ('Kanangurrwu' JU19)

(9) abanajukun [right arrow] yabanajuku-yu ('Ngadburriyingurriyi' JU17)

Note that, in the examples (7-9), the vocable is inserted after deletion of the final consonant of a preceding word. Coda deletion resulting in the sequencing of CV (consonant, vowel) syllables is a strong tendency in both Iwaidja song and speech (discussed further below).

In a few instances the inserted vocable consists of two syllables rather than one, as the following examples show.

(10) kirrimul [right arrow] irrimul-aya ('Jawina' JU14)

(11) ngaldalmaldangkayangkaj [right arrow] ldalmaldang-kayangka-yiya ('Nganbaldakaniny' JU09)

In other instances ya seems to be prosodically integrated into the word it appends to:

(12) jawina [right arrow] jawinya-ya ('Jawina' JU14)

Vocables as melodic carriers over whole metrical unit: In several instances, a vocable or set of vocables encompasses an entire metrical unit, either as the first half of a text line:

(13) ayaya nganbaldakani ('Nganbaldakaniny' JU09, Line 3)

(14) ayaya rrijumarludbu ('Jawina' JU14, Line 3)

or as a whole line consisting of two or more metrical units:

(15) angakakayaya ri ri ... ('Kudnuka ngartung' JU01)

(16) ayangakayaya ri ri ... ('Riwujbakba' JU12)

(17) uuu ... ('Kartbirljuju' JU10, 'Bujikad' JUll)

Vocable insertion in David Minyimak's 314 songs (JU22-32): An interesting case of metrically motivated vocable insertion is found in a set of songs composed by David Minyimak. Here a vocable we have transcribed as 'aa' is regularly inserted at the end of the first two words of each section, and two vowel vocables 'uu ii' (or similar) are regularly performed at the end of every verse (Figure 5 provides an example).

In this case, the first vowel vocable 'uu' harmonises with the last vowel of the preceding word, 'janamirrakbun'. In several other songs that share this tune a rounded vocable is pronounced in a more forward position if the final syllable of the preceding word contains a different vowel: its quality is closer to 'o' after 'ngaldalmalangkaj' (JU23) and 'yakaldadbarjan' (JU27), and to '6' after 'angmungulkbajiny' (JU24), and 'yanawurnyakbin' (JU29). The final 'ii' vocable is possibly a variant of 'ri' (the vocable that is repeated at the end of most other Jurtbirrk songs). Dropping the consonant onset of words beginning with/r/ is also common in speech (for example, the 3sg. masc transitive prefix 'ri' is mostly realised without its consonant onset).

The basic rhythmic structure of this song is as follows (each metrical unit is a 3/4 bar, with each line composed of two 3/4 bars):

* Line A + aa (2 bars)

* Line B (2 bars)

* Line B truncated (2 bars)

* Line C + aa (2 bars)

* Line D + uu + ii(2 bars)

[FIGURE 5 OMITTED]

All other songs in this set follow exactly the same ten-bar pattern, although sometimes the text-line repetitions are structured AAABC rather than ABBCD.

From Figure 5 it can be seen that melodically (as opposed to rhythmically) the 'aa' belongs to the following line rather than the preceding one, so that melodically the song can be grouped as"

* Line A (descent from D-flat to E-flat)

* aa + Line B + Line B truncated (descent from upper E-flat to A-flat, lingering on B-flat en route)

* Line C (descent from D-flat to E-flat)

* aa + Line D + uu + ii (descent from upper E-flat to A-flat, lingering on the final)

Rhythmic setting of words

In the Jurtbirrk song-set, regular patterns in the rhythmic setting of words can be identified, with variation according to rhythmic mode. As has been widely reported for other repertories of Australian song, there is a strong overall tendency for words to be set using short durations for syllables at the beginning of the word and longer durations for word-final syllables. The following discussion abstracts the syllabic rhythmic values from the melodic settings (that is, it does not take account of melisma, where a syllable is performed over more than one pitch in the melody).

Matching of words to metrical units in David Minyimak's 314 songs: In the set of songs just discussed, there is also considerable regularity in matching of words to the 3/4 metrical unit, as outlined in Table 6.

Note that all settings favour shorter notes at the beginning of the metrical unit and longer notes at the end of the metrical unit. It can be seen that there are two possible rhythmic settings of five-syllable words to the metrical unit: angmarranguldiny is set with an apparently longer second syllable, while janamirrakbun is set with an apparently longer third syllable (the usual setting for five-syllable words in this corpus). It is possible that this difference is due to underlying phonetic processes, such as the greater consonant duration of--ngm- in angmarranguldiny.

Similar patterns emerge for 9/8 songs, with the metrical unit of three clap-stick beats corresponding to fairly predictable rhythmic settings of a word (or sometimes, with words of two or three syllables, to two words). In 9/8 songs there tends to be more use of swung and syncopated rhythms, and therefore more variability in the precise combinations of the shorter notes and their alignment with the clap-stick beating. Table 7 shows that there is some evidence that particular composers or performers prefer slightly different settings of four-syllable words set to the three-beat metrical unit.

Note that in each case the final long syllable is of the same duration (a crotchet tied to a dotted crotchet). David Minyimak's version is the most commonly performed rhythmic setting, and can be found in performances by all singers. Reggie Cooper's rendition of yabaninga is unusual both within his repertory and within the whole Jurtbirrk corpus: it has a swung feel, as no syllable onset coincides with the second clap-stick beat. The third setting, by Ronnie Wandijak, results in a duple rather than triple subdivision of the clap-stick beat duration. This alteration is found throughout the Jurtbirrk repertory, but is especially prominent in performances by this singer.

Jurtbirrk songs: linguistics

Phonology

Syllable structure: Although the majority of Australia's Indigenous languages have consonant-initial word structure, Iwaidja is one of a minority of Australian languages that allows words to begin with a vowel, the name of the language itself providing an example. In addition, Iwaidja has many words containing syllable-final consonant clusters: for example, warrkbi 'man'. These are two ways, therefore, in which Iwaidja syllable structure diverges from the commonly posited universal preference for CV syllables. However, a comparison between the way isolated words are produced when spoken clearly (hyper-articulated) in Iwaidja, and the realisations of the same words in typical connected speech, shows a stark contrast in terms of syllable structure. There are phonological processes observable in connected speech that appear to move syllables closer towards conforming to the universally preferred CV structure. These processes are shown below in examples (18-21).

Unsurprisingly perhaps, these processes are also observable in song. If anything, however, the syllable structure of sung Iwaidja moves even closer to the CV pattern. This is brought about in two ways.

(18) Utterance-Initial Vowel Deletion

ang-mana-warlkbarrakany-mi-kbin [right arrow] a [n]manawalkparakanmikpin

2sg-FUT-old person-INCH-ITER

'you will grow old'

(19) Word-Final Consonant Deletion

bu-ma-ngung [right arrow] bumanu

3plA>3O-get-PST.HAB

'they used to gather it'

(20) Vowel Coalescence

yang-man-ara ang-mana-min [right arrow] jangmanajamanamin

AWAY:2sg-FUT-go 2sg-FUT-tell

'go and tell'

(21) Cluster Reduction

warrkbi man [right arrow] [wareg.sup.w]i

'man'

(22) Tap Epenthesis

yang-man-ara ang-mana-min [right arrow] jagmanajaramanamin

AWAY.2sg-FUT go 2sg-FUT tell

'go and tell' ('Yangmanara' JU10)

(23) nganbaldakaniny [right arrow] ngalbaldakani ('Nganbaldakaniny' JU09) alan > [right arrow] 'Jawani jukan' (JU02) wurruwarr [right arrow] ('Wurruwarr' JU08) kirrimul [right arrow] irrimu ('Ayunman wingalmu' JU18) wularrud [right arrow] ('Wularrud' JU32)

Firstly, the processes shown above are augmented by a fifth process not observed in speech, namely the insertion of an epenthetic tap, either retroflex or alveolar, between two vowels. Hence the phrase given in (20) above may have the alternative realization given in (22) when sung.

Secondly, at the other end of the syllable, word-final consonant-deletion applies more broadly than in speech. While in speech this process is mainly limited to word-final velar nasals, and palatal nasals following high front vowels, in song we see a broader range of deletions, shown in (23) below.

Alignment of clap-stick beats with consonant-vowel transitions of metrically strong syllables:

While other studies of Indigenous music traditions in Australia have suggested that the stress patterns of words are altered significantly when they appear in song texts (Hercus and Koch 1995; Strehlow 1971), in Jurtbirrk songs this is not the case. On the contrary, there is a strong tendency for the CV (consonant to vowel) transitions of metrically strong (stressed) syllables to align with clap-stick beats throughout the song corpus. (13)

Figure 6 shows a spectrographic representation of the first line of a rendition of the song 'Kurrana' (JU05). The text is glossed in Example (24) below.

(24) a'ralpa'ralpa 'juw-a-l'ak'palkpa be_quick! TOWARDS:2sg-IMP-reveal_thoughts

'quick, tell me what you're thinking'

In the spectrogram, clap-stick beats can be seen aligning with the CV transitions of metrically strong syllables in both words: the twin syllables ral in the reduplicated form a'ralpa'ralpa, and the syllable palk in 'juwal'ak'palkpa. (14)

The alignment of clap-stick beats with the CV transitions of metrically strong syllables has two ramifications for the analysis of prosodic structure in spoken Iwaidja. First, it supports the characterisation of Iwaidja as a 'stress accent' language (Beckman 1986), in which pitch accents align with metrically strong syllables, as opposed to a 'non-stress accent' language, where the intonation system is clearly oriented towards the chunking of speech into prosodic units, a well-known example being the accentual phrase in Korean (Jun 1998). Evidence for this in spoken Iwaidja comes not only from analysis of pitch accent placement (Birch 2000), but also from observations regarding variation in the quality of vowels, which tend to be peripheralised in accented syllables, and centralised or deleted in metrically weak syllables (Birch 2002a).

[FIGURE 6 OMITTED]

In song, however, the intonation contour normally associated with a spoken word or phrase is subordinated to a melodic contour that varies in the same dimension--that is, F0 (15) (pitch)--and according to different constraints. In the case of the Jurtbirrk songs, although the usefulness of F0 target alignment as a diagnostic of metrical strength is greatly reduced for this reason, it turns out that clap-stick beat alignment provides a reliable substitute diagnostic. An analysis of the metrical structure of words in the Jurtbirrk song corpus based on the alignment of clap-stick beats with CV transitions produces much the same results obtained for the spoken language using the diagnostics mentioned above.

The second ramification of the alignment of clap-stick beats with the CV transitions of metrically strong syllables in song is that it reflects the universal tendency for syllables to be structured as sequences of a consonant followed by a vowel (CV), as opposed to the inverse order, that is, a vowel followed by a consonant (VC). The songs constitute, in a sense, a naturally occurring analogue of experiments that have been conducted to find the P-centre (perceptual centre) of syllables (Marcus 1981). In these experiments, when subjects were asked to synchronize clicks with syllables, it turned out that the clicks were aligned at a certain point, known as the P-centre, close to the CV transition of the syllable. The corresponding alignment of clap-stick beats with CV transitions in Jurtbirrk songs adds support to evidence from phonotactics and connected speech processes (16) that syllabification in Iwaidja patterns according to the near-universally attested 'onset first' principle.

Word boundaries marked by melodic (F0) turning points: As mentioned in the previous section, in spoken Iwaidja, when two words or phrases are juxtaposed with no intervening pause, and where the first word or phrase ends in a vowel and the second begins with the same vowel, these two vowels coalesce to form a single long vowel, thus masking the edges of the words involved. Despite their deletion in the segmental layer in such cases, however, word boundaries frequently remain perceptually salient due to a feature of the Iwaidja intonation system, in which the right edges of words, or word-sized units, tend to align with low F0 targets or turning points in an intonation contour. (17)

In cases of vowel coalescence across a word boundary, therefore, a low F0 target, if present, provides a clear acoustic diagnostic of the location of the boundary, tending to be aligned as in Figure 7. Low targets or turning points align in this way even if the initial syllable of the following word or phrase is metrically weak (unstressed), suggesting that it is the boundary of a word or phrase that is being marked, rather than the edge of a smaller prosodic unit such as a stressed or accented syllable.

In the Jurtbirrk song corpus, there appear to be examples of this same phenomenon, though in this case incorporated into, and constrained by, the melodic contour of the song. Figure 8 shows a spectrogram and F0 trace of an excerpt from Line 1 of a rendition of 'Yangmanara' (JU03). The line from which the extract is taken is glossed in Example (25) below.

(25) jan-'man-a-ra an-'mana-min 'fumun AWAY:2sg-FUT-go 2sg-FUT-tell 3sg.OBL

'got and tell her'

The spectrogram shows a long low coalesced vowel spanning the boundary between the words jan'manaia and an'manamin. Beginning at the left edge of the diagram, the F0 trace shows the pitch being maintained at the fourth scale degree (A) before descending to the third (G) at a point roughly simultaneous with the second visible clap-stick beat.

The beginning of the next pitch transition, however, a rise from G to B, is not aligned with a clap-stick beat. Rather, it appears to be aligned with a word boundary in the song text, anticipating the clap-stick beat by some 0.126 seconds. As in speech, although the low vowels of the two words have coalesced, leaving no trace of a word boundary in the spectrogram, the location of a turning point in the F0 contour is consistent with the predicted location of a word edge. F0 rises through the onset-less first syllable of the word an'manamin, before reaching the next melodic target (B) at the clap-stick beat, which has its typical alignment with the first metrically strong syllable of the word.

This example shows that, although the usefulness of intonational features as diagnostics of prosodic structure is inhibited by the subordination of the intonation contour to the melodic contour, as suggested in the previous section, certain key elements of spoken intonation in Iwaidja may remain intact when a text is sung.

Vowel modification: In a quantitative study of vowel quality in spoken Iwaidja (Birch 2002b) (18) it was found that vowels are centralised in relation to the mean when they occur in frequently occurring lexemes or constructions, and when they occur in metrically weak syllables. (19) On the other hand it was found that vowels are peripheralised when they occur in syllables aligned with pitch accents. As the results of other studies (for example, Recasens 1985) predict, the study also found that the low central vowel a exhibited a greater degree of variation than the two high vowels.

[FIGURE 7 OMITTED]

In the Jurtbirrk texts, an examination of vowel quality reveals an almost inverse set of tendencies to those observed in speech. The low central vowel exhibits far less centralisation than in spoken Iwaidja, whereas the two high vowels are lowered and centralised to a greater extent, the front high vowel typically being realised as [epsilon], and the high back vowel typically being realised as e or a. In addition, it would seem that the processes of centralisation and lowering of vowels in song are not attributable to the factors influencing them in speech.

It is particularly illuminating to compare the formant values for vowels occurring in lengthened syllables in song with their counterparts in hyperarticulated emphatically accented syllables in spoken Iwaidja. Figure 9 shows a typical realisation of a long high front vowel in the Jurtbirrk corpus.

The lengthened vowel in the second syllable, though phonologically high, having typical F1 and F2 values of around 420Hz and 2600Hz respectively under hyperarticulation, here has F1 and F2 values of 825.42Hz and 1795.54Hz. This gives the vowel the acoustic profile of a far lower and more central vowel, here transcribed as e, and roughly equivalent to the long vowel in the English word air.

Figure 10 shows the word 'jumun '3sg.OBL' extracted from the same performance. Hyperarticulated tokens of the low back vowel in spoken Iwaidja have typical F1 and F2 values of 470Hz and 920Hz respectively. The realization of this vowel in the lengthened second syllable of 'jumun, transcribed here as a, is more typical of a hyperarticulated low central vowel in Iwaidja, though slightly further back. It has an F1 of 904.35 and an F2 of 1469.44Hz.

In contrast to the two high vowels, which, when they occur in song, do not appear to exhibit the formant values typical of their hyperarticulated spoken language counterparts, the low central vowel exhibits a tendency to be hyperarticulated where it would typically be centralized or deleted in the spoken language.

[FIGURE 8 OMITTED]

[FIGURE 9 OMITTED]

An example of this occurs in the metrically weak syllables of the reduplicated form a'ralpe'ralpa (Figure 6 for reference). The two pre-retroflex low vowels in this word are commonly centralised to schwa or deleted in spoken Iwaidja, resulting in the variant realisations e'ralpe'ralpa and 'ralpe'ralpa. In this sung version, the first pre-retroflex low vowel has F1 and F2 values of 758.74Hz and 1546.18Hz respectively, while the second is slightly lower and further back with values of 804.55Hz and 1442.78Hz. In both cases, then, and in spite of the fact that they have short durations, the vowels are realised as typical low central vowels, showing no signs of centralisation.

A probable motivation for the differences in vowel variation tendencies in sung, as opposed to spoken Iwaidja, is the need to maximise voice projection when singing. The high F1 values for all vowels suggest that the jaw is lowered in all cases. This suggests in turn that, as in the Western classical tradition where singers are trained to modify vowels in order that their unamplified voice may project above the level of an orchestra during performance, Jurtbirrk singers adopt similar techniques so that their voice can be heard above accompanying instruments and ambient party noise. Hyperarticulated high vowels with their constricted, more consonant-like articulations, are far from optimal in terms of maximising voice projection, and are therefore likely to be avoided in such settings.

[FIGURE 10 OMITTED]

Grammar

Jurtbirrk songs are sung in ordinary language, drawing on the full richness of Iwaidja grammar. Indeed, some grammatical devices are employed with particular frequency, so much so that our study of Jurtbirrk has advanced our understanding of one central part of Iwaidja grammar, as we discuss in this section.

A characteristic feature of Iwaidja grammar is that all verbs take their pronominal prefix from one of three directional sets: neutral, towards, and away. Examples (26-28) illustrate their use.

(26) a-ra 3pl-go

'They go, they are going along.'

(27) ijb-ara AWAY: 3pl-go

'They go, they are going away.'

(28) ayuw-ara TOWARDS: 3pl-go

'They come, are coming.'

Directionals are not especially unusual in Australian languages--for example, Warlpiri places the directionals -rni/-rnu 'hither; towards' and -rra 'thither: along, away, across' right after its verbs. But whereas in Warlpiri a single form can be combined more or less unchanged with all subjects and objects, in Iwaidja they get tangled up in a paradigm or multidimensional table where it is hard or impossible to say which sound(s) contribute which meaning(s). More precisely, every verb in Iwaidja takes a prefix encoding the following semantic dimensions:

(29) Subject: person x number (x gender just for 3sgSubj>3Obj)

(If transitive--Object: person x number) Mood: (indicative versus irrealis versus imperative) Direction: neutral versus towards versus away

For example, the prefix iarrubana- means 'we, in the future, will act on him or her, in an away direction'. So if you said 'we will take him or her away' you could say jarrubanawilimang, combining it with the verb wilimang 'take'.

Multiplying the values in (29) means that each transitive verb has hundreds of possible prefix combinations that essentially need to be learned by heart, since you can't just put together a single bit (or 'morpheme') for each meaning. Table 8 compares some neutral and away forms; the right hand column gives the best guess about what the 'away' morpheme is, based on the difference between the two other forms.

Since the interactions are complex and non-predictable, many distinct forms need to be learned and documented, and obtaining a full set of paradigms is one of the slow chores of documenting--or completely learning--a language like Iwaidja.

To make matters worse, it appeared until recently that either some combinations were so unnatural that it was impossible to make up a context where you could sensibly ask for an example, or perhaps that some forms just didn't exist. (And it is not unusual that a language may not bother to have forms for very rare combinations. After all, children only learn a language by hearing it, and if they never hear rare forms, how would they know what they are ?) The otherwise quite detailed description of Iwaidja by Pym and Larrimore (1979) has gaps for many combinations: all the 'away' forms involving a first or second person object (that is, of the type 'he > me away; I > you away', and so forth), and all the 'towards' forms involving interactions between first and second persons ('I > you towards', 'you mob > us towards' and so on). Based on three to four years of fieldwork in 1973-5 and 1977, they stated (Pym and Larrimore 1979:94): 'According to the present data a complete set of directional-person-number prefixes is not in use with the transitive verb. Only those with third person objects are in common use.'

In recent years, two of us (Evans, Birch) thought we might have more luck, and again tried eliciting some of these forms, trying to get consultants to describe situations like 'you gave it away to me', both in Iwaidja and the closely related Ilgar (now extinct). Speakers of these languages would just say 'you gave it to me'. Perhaps, we concluded, there was no need to express the meanings 'towards' or 'away' in interactions involving the speaker and the hearer, since the direction of transaction is clear anyway. To our language teachers' relief, we gave up trying to elicit these cells of the paradigm.

However, once we began recording and analysing Jurtbirrk songs, there was a new twist.

As it turns out, Jurtbirrk songs make incredibly frequent use of directional forms: of 32 songs in the collection, 22 have at least one directional form. Twenty include at least one 'away' form, and three include at least one 'towards' form; one song has both. Considering that the major themes are intimacy (getting your lover towards you) and abandonment or longing (your lover has gone away from you, or you are longing for your lover who is away) the basic reason is clear. Sometimes the directionals are used in their literal meaning of spatial movement, and sometimes in more metaphorical senses. In fact, the range of meanings that directionals express is more nuanced than this:

(a) moving away / toward

(b) moving along (away form)

(c) continuous or ongoing action (away form; like English talking away)

(d) action carried out at distance versus close (for example, cooking at distant versus close fire)

(e) covering other gaps in the paradigm, with verbs whose object is filled by something else (for example, a thing given or cooked), so that 'away' can suggest 'to/for you' and 'towards' can suggest 'to/for me'. (20)

Here are some translated examples from the Jurtbirrk collection illustrating these various uses:

Towards:

* singer moving toward location of song ('Kuluduk' JU25)

* addressee should reveal thoughts (to speaker) before the moon comes up (towards) ('Kurrana' JU05)

* warning (to secret lover?) not to enter (towards) in case others see ('Yinang birta' JU07).