The Language of Wangga
3. CHAPTER 3
3. THE LANGUAGE OF WANGGA
This chapter analyses the language of wangga song texts and shows a much more complex picture with regard to the differences between song language and everyday spoken language than is generally accepted. The following statement by Dixon is typical of generalisations about such differences:
Songs that involve just everyday vocabulary tend to have quite simple grammar. There is minimal use of affixes—often in marked contrast to the everyday speech style—with, for instance, case-endings sometimes being omitted from nouns. In many languages songs just involve simple sentences, eschewing complexities. (Dixon, 1980, p 56)
I have already shown, through a linguistic analysis of Marri Ngarr lirrga song texts, that lirrga songs reveal a significantly greater degree of morphological and syntactic complexity than survives in spoken texts (Ford, 2006). In this chapter, I will show that, like lirrga, wangga song texts in the five language varieties in which wangga is sung contain significant complexity in their grammar and lexicon, and that they preserve features no longer present in everyday language. Some features absent in wangga that are present in everyday spoken language might be explained by the constraints of metre and poetry.
Most wangga texts come from a period when the languages in which they are sung were viable, when the songmen and their audiences had command of more than one of the Australian Indigenous languages of the Darwin-Daly River region. Most songmen sang in only one of the Daly languages, but some songmen used more than one language in a single song: Lambudju alternated languages at the level of text phrases (chapter 7, track 11); Mandji alternated them at the level of vocal section (chapter 6, track 8). Today, the speech communities that produced wangga are no longer viable. Thus wangga texts are an important source of information about the grammar and lexicon of morphologically complex, severely endangered Australian Indigenous languages.
Wangga is sung in five language varieties from three distinct languages. These languages are:
- Batjamalh (the language of the Wadjiginy), still spoken at Belyuen (see map) but with fewer than half a dozen fluent speakers;
- Emmi (the language of the Emmiyangal people) and its closely related dialect Mendhe (the language of the Mendheyangal people) with only a handful of fluent speakers at Belyuen, One Mile Dam and Knuckey's Lagoon (both in Darwin) and at Balgal outstation opposite the Peron Islands;
- Marri Tjavin (the language of the Marri Tjavin people) and its closely related dialect Marri Ammu (the language of the Marri Ammu people) with perhaps a dozen fluent speakers at Wadeye, Perrederr and at Knuckey's Lagoon in Darwin.
The traditional territories of these five language varieties abut each other, with Batjamalh the most northerly language, followed, along the coastal strip to the south of the Daly River mouth, by Emmi, Mendhe, Marri Ammu and finally Marri Tjavin. Emmi-Mendhe and Marri Tjavin-Marri Ammu belong to the Western Daly group of languages, a low-level linguistic subgroup (Dixon, 2002, p 675). The linguistic relationship of Batjamalh is less certain. Loans from Emmi-Mendhe into Batjamalh suggest that Batjamalh is a relative newcomer to the area, but even so it is likely that the Wadjiginy and their language have been in the Daly River region for several hundred years.
All five language varieties used in wangga are by now severely endangered. None has more than a dozen fluent speakers, and most current songmen have only imperfect command of Batjamalh. This is because of relatively recent but thoroughgoing language shifts.
As pointed out in chapter 1, the past one hundred and forty years have seen the migration of most speakers of Batjamalh and Emmi-Mendhe from their traditional territory around the mouth of the Daly River to Darwin, founded in 1867. The first recorded evidence of the presence of the Wadjiginy at Point Charles, on the west coast of Darwin Harbour, is dated 1885 (Ford, 1990). In 1937, the Wadjiginy and their Emmiyangal and Mendheyangal affines were resettled across the harbour from Darwin at Delissaville, now known as Belyuen. Today, everyone aged fifty and under at Belyuen speaks light Kriol (Sandefur, 1991) as their first language.
More recently, speakers of Marri Ammu and Marri Tjavin were brought into the Catholic mission founded in 1935 at Port Keats (known since 1979 as Wadeye). Marri Ammu and Marri Tjavin are related to the neighbouring inland language Marri Ngarr and its coastal dialect, Magati Ke. All four language varieties are generally known as the Marri languages, and their speakers are known as members of the Marri language group. Marri elders tell how, as a result of intimidation by the more numerous Murriny Patha people, who own the land on which the mission stood, they started to speak Murriny Patha as their main language (Ford & Klesch, 2003). Today, there remain few fluent speakers of the Marri languages, and an unpublished sociological survey carried out in 2000 by Ford, Kungul and Jongmin showed that everyone at Wadeye aged fifty and under speaks Murriny Patha as their first language (Ford, Kungul, & Jongmin, 2000). Those whose mother tongue belonged to the Marri subgroup found this language shift expedient and relatively easy because, despite thoroughgoing lexical differences, the sound system and grammatical organization of Murriny Patha are roughly similar to those of the Marri languages.
Overview of languages in which wangga is sung.
The languages in which wangga is sung have similar, but not identical, phoneme inventories. These are relatively simple. They all have a voicing contrast in stops and five series of stops and nasals. They all have two apical series of stops /t/, /d/, /rt/, /rd/, and all have two laminal series of stops (lamino-palatal /tj/, /dj/ and lamino-dental /th/, /dh/) except for Batjalmalh, which has only a lamino-palatal series. All three languages have an apical retroflex fricative /rz/. In addition Emmi-Mendhe and Marri Tjavin-Marri Ammu have a bilabial fricative /v/ and Marri Tjavin-Marri Ammu has an additional fricative, the lamino-palatal /sj/. All have two peripheral series of stops: labial /p/, /b/ and dorso-velar /k/, /g/. They all distinguish six nasals /n/, /rn/, /ny/, /nh/, /m/, /ng/, except for Batjamalh, which lacks /nh/. They all have two rhotics (apical /rr/ and retroflex /r/); three laterals (apical /l/, /rl/ and laminal /lh/), and two glides (lamino-palatal /y/ and a labial-velar /w/). They all have the four phonemic vowels /i/, /e/, /a/, /u/, but in addition Batjamalh has a fifth high front rounded vowel /ü/ and Emmi-Mendhe has a fifth mid unrounded vowel /ö/. These inventories are given in full below in discussion of the sound systems of the relevant languages.
Each of these languages contains a lexicon of several thousand words (Ford, 1997) and in preparation). Each language uses sets of nominal and verbal classifiers to order the world around it, and contains an elaborate set of kin-terms by which it orders relationships between individuals. The number of classifying verbs ranges from ten in Batjamalh to thirty in each of Emmi-Mendhe and Marri Tjavin-Marri Ammu. Each classifying verb may co-occur with one or more of hundreds of co-verbs.
These are morphologically complex, polysynthetic languages. This means that they contain long words consisting of ordered morphemes. Some of these morphemes are so old that they have fused, and thus convey more than one piece of meaning. Verbal affixes indicate, often redundantly, the person, number and gender of Subject and Object, and verbs inflect to show tense, aspect, mood and modality. Nominals are case-marked. A rich system of enclitics indicates direction, aspect, purpose and illocutionary force.
Analysis of each of these languages contains the discrete word classes of nominals (nouns, pronouns, adjectives, deictics), verbals, modifiers (adverbs) and particles. Except for enclitic particles, which may attach to words of any class and are often attracted to a clause-initial host, each of these word classes behaves differently from the others and takes different affixes.
Syntactically, each of these languages is head-marking, in that its verbs inflect for person, number, and gender. Each language is also dependent-marking, in that its nominals are, as in many Australian languages (Dixon, 1980, p 294), case-marked for core, peripheral and local relations. All sentient beings—including humans, living creatures and the land itself—qualify as Subjects.
Verbs are formally marked as transitive or intransitive by means of pronominal prefixes, which specify the person and number of the Subject and, in the case of transitive verbs, the Object of the verb. In Batjamalh, Subject and Object are fused in a portmanteau prefix. In the other languages, Subject and Object are marked by discrete prefixes. In Batjamalh, third personal minimal prefixes are also gender marked.
In each of these languages, the number system marked on verbs and free pronouns is most accurately described as Minimal versus Augmented, rather than Singular versus Plural. In each, there is a Minimal stem for: I, you (one person), you (one person + I—speaker and addressee) and he/she; and an Augmented stem for we/us, you (more than one person) and they/them (glossed MIN and AUG respectively). Each language has an affix meaning 'plus one', which suffixes to a Minimal stem to augment the stem by one.  Each language has another affix meaning 'plus more than one', which suffixes to an Augmented stem to specify numbers greater than three.
Each of these languages has a rich system of bound pronouns describing how the action/event denoted by the verb affects a human Object, and the degree to which the person affected has control over what is happening to them. Free and bound pronouns make a distinction between including or excluding the addressee (inclusive/exclusive).
Each language employs verbless equational clauses and verbal clauses to state propositions. Verbs inflect for Realis mood to express statements whose reality can be vouched for by the speaker, and for Irrealis mood when the speaker refers to events, states or actions that he or she has not witnessed. These include things yet to happen, hearsay and hypotheticals, indeed anything whose reality the speaker cannot vouch for.
Each language has simple and complex clauses including embedded clauses. These may take the form of headless relative clauses, which omit the antecedent to the relative clause. Each language also contains serial constructions, which are conjoined clauses containing a main verb, followed by a simple intransitive verb signifying the physical orientation of the co-referential Subject, and/or adding aspectual information to the notion expressed by the verb. This aspectual information tells us whether the action/event/state expressed by the verb is habitual, repeated, deliberate, temporary or permanent. Both verbs share specification for tense (Past/Non-past), mood (Declarative/Imperative) and modality (Realis/Irrealis).
Semantically, each of these languages incorporates body-part nominals into a verb phrase, or noun phrase to signify any entity of the same shape. So, Batjamalh mive, like Emmi-Mendhe and Marri Tjavin-Marri Ammu miri, 'eye', stands for a face, a person, a window or door, a seed, a hearth or a fire. Batjamalh dawarra, like Emmi-Mendhe mari, and Marri Tjavin-Marri Ammu marzi, 'belly,' stands for the interior of a boat or car, or the curve of a beach. In everyday spoken language, body parts are metaphors for emotion: the belly represents anxiety, the back represents laughter and the head represents embarrassment.
In Emmi-Mendhe and Marri Tjavin-Marri Ammu, classifying verbs for the most part co-occur with semantically appropriate co-verbs. While some pairings appear anomalous, this is inevitable in a system where co-verbs require classifying verbs to specify the person, number and gender of the Subject and Object, as well as the tense, aspect, mood and modality of the action/state/event denoted by that co-verb. In Batjamalh, this system applies to some verbs, but by no means all, because many verb stems inflect for person, number, gender, tense, aspect, mood and modality without requiring a classifying verb to do this.
Each of the song repertories discussed in this book contains, to a greater or lesser extent, words that are not part of normal spoken language. In transcription we gloss these as 'song words' (SW) and in identifying such words, we are reliant on songmen and other informants, who usually say that they are 'just for song.' In many cases vocables of this sort are identified as 'spirit language.' For example, Tommy Barrtjap told us that when he received songs from Wunymalang ghosts in dream, they were initially entirely in the language of the Wunymalang, and that it was his responsibility, as songman, to 'turn over' the songs into human language. In many of his songs, only a portion of the text is rendered in human language (in this case, Batjamalh), while the remainder of the song is left in Wunymalang language, which we, like native speakers, hear only as a string of unintelligible vocables.
Contrary to the experiences of Dixon and Koch (1996, pp 26-34), our consultants have, over several decades, steadfastly resisted our attempts to segment or assign meaning to individual song words. Moreover, there was complete clarity about what were song words and what were not, even in cases where the form of the song words bears a resemblance to words in mundane register. For example, the phrase mele nele—which could mean 'older brother, for him'—occurs in one of Mandji's songs within a section of the song that otherwise contains only song words (chapter 6, track 1, text phrases 1 and 2). Consultants were adamant that in this context mele nele are song words and not words in Mendhe. In one of Barrtjap's songs, 'Kanga Rinyala Nga-ve Bangany-nyung,' (chapter 4, track 5) the phrase kanga rinyala tantalisingly seems to contain the Batjamalh word for melody, rinya, yet attempts on our part to gloss it as Batjamalh rather than song language were strongly resisted by the songman himself.
Spirit language resists linguistic analysis, because fluent speakers of the languages under focus insist that song words are not generally understood by living humans. Moreover, song words are generally CVCV sequences that do not correspond to words in everyday register. For this reason, I will not be dealing with song words in any detail in this chapter, although in the following discussion of Batjamalh, I will discuss some of the difficulties of transcribing such words.
Nevertheless, it is clear that the proportion of song words to words in everyday language differs from singer to singer. Of the two Batjamalh-speaking songmen, for example, Barrtjap uses a much higher proportion of song words than does Lambudju, and so do the other two Belyuen songmen, the Mendheyangal songman Jimmy Muluk and the Marri Ammu speaker Billy Mandji. The Marri Tjavin and Marri Ammu wangga from Wadeye, on the other hand, use almost no song words at all. Although there is some overlap, the song words in the Emmi-Mendhe and Marri Ammu-Marri Tjavin wangga of Muluk and Mandji generally differ from those in the Batjamalh wangga of Tommy Barrtjap.
I begin my detailed analysis of wangga song language with Batjamalh, not only because the earliest wangga in our corpus were composed in Batjamalh, but also because its lexicon and its verbal and nominal morphology differ significantly from those of the other languages in which wangga is sung. That Batjamalh is the newcomer to the area is shown by its fossilised and limited use of nominal, verbal classifiers and body part noun incorporation, and, as Evans (1989) shows, Irrealis morphemes. All of these are borrowed from the neighbouring Marri languages. The two Batjamalh wangga repertories were composed and performed by Wadjiginy songmen Barrtjap (chapter 4) and Lambudju (chapter 7).
The sound system of Batjamalh contains fewer consonant phonemes than those of the other languages in which wangga is sung. As table 3.1 shows, its twenty-three consonants include five full series of voiced and voiceless stops: two apical, one laminal, and two peripheral, with nasals to match. Batjamalh lacks the full lamino-dental series of stops and nasals found in Marri Tjavin-Marri Ammu, and has only one, rarely occurring, phonemic fricative, compared to the two in Emmi-Mendhe and the three in Marri Tjavin-Marri Ammu. Like the other Daly languages in which wangga is sung, Batjamalh has three laterals (two apical and one laminal), two apical rhotics and two semivowel continuants.
As pointed out in the previous section and set out in table 3.2, Batjamalh has five phonemic vowels, including the rounded front vowel ü.
Table 3.1 Batjamalh consonant phonemes
Table 3.2 Batjamalh vowel phonemes
Tables 3.1 and 3.2, listing Batjamalh consonant and vowel phonemes, may be compared with the inventory of Emmi-Mendhe consonant and vowel phonemes shown in tables 3.3 and 3.4, and the inventory of Marri Tjavin-Marri Ammu consonant and vowel phonemes shown in tables 3.6 and 3.7.
All Batjamalh phonemes are represented in the wangga song-texts, except the apico-postalveolar (retroflex) fricative and lateral. The absence of these phonemes in wangga is not surprising, as both carry a low functional load in everyday speech. In my extensive Batjamalh corpus, /rz/ occurs only in the lexemes marzanmarzan, 'barnacle,' and mürza, 'star,' and its compounds mürzarak, 'sky,' and mürzamedjem, 'egg,' while /rl/ occurs only in the lexemes durl, 'cranky,' durlk, 'whale; Dreaming,' kurluk, 'blind,' and barndarla, 'stringybark.'
So as to represent as faithfully as possible the actual sounds sung, and to render the text as accessible as possible to the general reader, our transcriptions of wangga texts are written in a sub-phonemic orthography that shows the allophonic variants of phonemes.  For instance, the voiced bilabial stop phoneme /b/ is often realised as the allophone [v]. We have therefore made a decision to represent /b/ as [v] when it occurs between vowels. This can be seen in text phrases such naya rradja bangany nye-ve, 'Naya rradja. You go for a song' (chapter 4, track 16, text phrases 1-3) and yagarra yine nga-ve-me-nüng, 'Yagarra! What have I come to do?' (chapter 4, track 7, vocal sections 2 and 3, text phrase 2; track 22, text phrase 3).
A second example of our use of sub-phonemic orthography occurs where consonants show allophonic assimilation to a preceding or following consonant. A frequently occurring example of the former is where the initial consonant of the dative suffix ‑nung assimilates to the final consonant of the preceding morpheme bangany (song) to produce bangany-nyung (chapter 4, track 1; chapter 7, track 1). An instance of allophonic assimilation to a following consonant is found where the final apico-alveolar nasal of the pronominal prefix ngan- is realised as retroflex before a following stem-initial retroflex stop, see for example ngan-rdut-mene-ng represented as ngarn-rdut-mene-ng (chapter 7, track 10, text phrase 1b).
A third example of our use of our use of sub-phonemic orthography concerns vowel harmony, which occurs when a vowel assimilates to a preceding or following vowel. An example of the former occurs where the initial vowel of a suffix assimilates to the vowel quality of the final vowel of the preceding morpheme, as when the suffix ‑nung becomes ‑nüng after me, as in yagarra yine nga-ve-me-nüng (chapter 4, track 7, vocal sections 2 and 3, text phrase 2; track 22, text phrase 3) or after ‑djü, as in nga-p-pindja-ng nga-p-puring-djü-nüng (chapter 4, track 7, vocal sections 2 and 3, text phrase 3). An example of assimilation to a following vowel occurs in bangany nye-bindja-ng nya-mu, 'Sit and sing a song!' (chapter 7, track 17, text phrase 1) where there is vowel harmony affecting the final vowel of each pronominal prefix nyV. The underlying vowel of this prefix is realised as [e] before -bindja because the first vowel of -bindja is a front vowel but is realised as [a] before -mu because -mu contains a back vowel. Thus we write nye-bindja in the first case, but nya-mu in the second.
Transcribing the vocables in wangga song-texts presents particular difficulties, because there are no equivalents in everyday language against which to check them. We write them in the same sub-phonemic orthography as the rest of the wangga texts. They are, however, inherently unstable as a result of the distortion brought about by the airflow of singing, which is typically very strong at the start of a breath phrase and weak towards the end. This distortion blurs the distinction made between sounds produced in the same part of the mouth (Marett & Barwick, 1993, p 21). For example, if the apico-alveolar rhotic /rr/ is, for this reason, not accurately articulated, or pushed back in the mouth, rradja can sound like adja, dadja, rzadja or even gadja.  Similarly, if the lamino-palatal stop /dj/ in the same song word rradja is not accurately articulated, it can be lenited to rraya or fricativised to rrasja. For the same reason, nasals are more clearly articulated with the relaxation of breath towards the end of a breath-phrase, and conversely lightly articulated at the beginning of a breath-phrase. So, the first naya in text phrase 1 of each verse of 'Naya Rradja Bangany Nye-ve' (chapter 4, track 16) can sound like aya, and the quality of the nasal in mayave is hard to distinguish. Vowels tend to close and come forward and higher as the breath flow weakens, so that [a] becomes [e]. For this reason bangany-nyaya is sometimes realised as bangany-nyaye and yagarra becomes yakerre.
The wangga texts contain examples of all Batjamalh's word classes.
A well-formed Batjamalh sentence need contain no overt nominals. Batjamalh wangga song-texts, however, contain a higher proportion of overt nouns than is common in everyday speech. Barrtjap's 88 clauses contain 49 nominals and Lambudju's 46 clauses contain 56 nominals. Some of these nominals represent new information, some are repetitions, some are body-part nominals incorporated into the verb phrase, and others form verbless equational clauses stating the topic of the song-text, as in anadadada bangany-nyaya nga-bindja-ya '"Anadadada"' is the song I'm singing' (chapter 4, track 26, text phrases 1-2).
A wide range of nominals appears in the wangga texts. These include: proper nouns, either personal names such as Tjerrendet (chapter 7, track 15, text phrase 1) and Mangalimba (chapter 4, track 25, text phrase 1), or place names such as Badjalarr, 'North Peron Island' (e.g. chapter 7, track 1); common nouns such as winmedje, 'oyster' (chapter 7, track 1, text phrase 6; track 14, text phrase 1) or kurratjkurratj, 'channel-billed cuckoo' (chapter 7, track 13, text phrases 1-4); and kin-terms such as balhak, 'older brother' (chapter 7, track 10, text phrase 1b), ngaradja 'daughter' (chapter 4, track 1,text phrases 7-9) and nedja, 'son' (chapter 4, track 20, text phrases 1-4).
An example of an adjective in Badjamalh wangga texts is munguyil-malang, 'fast-paddling' (chapter 7, track 10, text phrase 3), which is derived from the noun munguyil, 'paddle,' by adding the suffix ‑malang, '-FUL.' Everyday spoken Batjamalh contains several adjectives produced by adding ‑malang, such as bwikmalang 'bony', from bwik, 'bone' + -malang, '-FUL.'
Although gender and number-marked deictics occur in everyday Batjamalh to position referents in relation to the speaker, the only deictic to occur in Batjamalh wangga is tjidja, 'this man near me' (chapter 4, track 23, text phrase 5 and chapter 7, track 15, text phrases 1-4). Other wangga repertories use deictics more freely.
Each nominal is case-marked to show its relation to other nominals in the sentence. As in everyday Batjamalh, nominals in wangga texts carry case-affixes encoding core syntactic, peripheral syntactic or local relations. As a core grammatical relation, a nominal in Absolutive case, which carries zero case-marking, may function as the Subject of an intransitive verb, as in Tjerrendet-maka  ka-ngadja 'Tjerrendet has gone back' (chapter 7, track 15, text phrase 1). A nominal in Absolutive case may also function as the Object of a transitive verb, as in winmedje ngan-dji-nyene ngami, 'I am sitting eating oysters' (chapter 7, track 14, text phrase 1).
When first introduced into the discourse, a noun or pronoun case-marked for the core syntactic role of Agent of a transitive verb carries the affix -garrang, which is realised as -karrang after a host-final voiceless stop, as in malvak-karrang-maka ngarn-rdut-mene-ng ka-bara, 'Malbak has gone and left me behind' (chapter 7, track 10, text phrase 1).
Batjamalh wangga texts also contain nominals case-marked to show dative, causative and instrumental relations. For example, a nominal is marked as a target or goal with the dative case-affix -nyung (e.g. chapter 4, track 1; chapter 7, track 1).  The causative case-affix -maka denotes the reason for an action, event or state. So, for instance, rak badjalarr-maka means 'for the sake of a song for my ancestral country, North Peron Island' (chapter 7, track 1, text phrases 1-5). The instrumental case-affix -djene marks the floating log as a tool in ngawardina-djene (chapter 7, track 10, text phrase 4).
Batjamalh has two ways to show the possessive relation. The most common in everyday speech is to affix the possessive affix -bütung to the possessed noun. The alternative is to juxtapose possessed and possessor nouns, as always happens in Marri-Tjavin and Marri Ammu. -bütung does not occur in Batjamalh wangga. When Barrtjap sings rak-pe ngadja, 'my eternal country,' however, possessed (rak-pe) and possessor (ngadja) nominals are juxtaposed (chapter 4, track 12, text phrases 1-4).
In everyday Batjamalh, verbs may incorporate body-part nominals that represent a part of the Object (metonymy). In tjendabalhatj mive-maka nyen-ne-ne kanye-djanga, 'Tjendabalhatj, they saw you standing there' (chapter 7, track 16, text phrases 1-4), mive, 'eye,' stands for the whole person called Tjendabalhatj. Similarly, müng ya-mara 'catch him up' literally means 'kick his arse (müng)' (chapter 7, track 10, text phrase 3).
As mentioned above, body-part nominals may also be metaphors for any entity with the same shape as the body part. In Lambudju's wangga, nguk in nguk ka-maridje-ng ka-yeve, 'he is lying with one knee bent over the other’ (chapter 7, track 10, text phrase 2) is the reduced form of the everyday word mirranguk 'knee.' While in this case the reference is to the actual body part, the knee, in everyday Batjamalh nguk may also signify 'pandanus fruit,' which happens to be shaped like a kneecap. Similarly, dawarra 'belly,' is used to refer to any concave entity. Features of the landscape are commonly referred to as the body-part whose shape they share. So, in Barrtjap's wangga, dawarra wagatj, 'belly of the beach,' means the hollow curve of a beach, as in yagarra dawarra wagatj nga-bindja-ng ngami ni 'I was sitting in the curve of the beach singing "ni"' (chapter 4, track 7, vocal section 1, text phrase 3).
Batjamalh wangga texts reveal copious evidence of verbal and non-verbal clauses. Of the 103 clauses that make up Barrtjap's wangga corpus, 53 (51.45%) contain verbs. Of the remaining 50 non-verbal clauses, 38 (76%) consist of the free particle yagarra (which we never translate, although it literally means 'oh, no!'), and the rest are equational clauses, such as nedja tjine rak-pe, 'Son, where is my camp/eternal country?' (chapter 4, track 20, text phrases 1-3).
Lambudju's wangga contain a slightly higher proportion of verbal clauses. Of the 62 clauses, 47 (75.8%) contain verbs. In the remaining 15 clauses, there are 6 elliptical constructions. In track 1 of chapter 7, for example, the causal noun phrase + dative noun phrase, rak badjalarr-maka + bangany-nyung, 'for my ancestral country, North Peron Island + for the sake of a song' is repeated five times with no main verb, thus giving, 'for the sake of a song for my ancestral country, North Peron Island.' In his spoken gloss, however, Lambudju supplies the missing verb nga-bindja-ng, 'I am singing.'
There are two types of Batjamalh verb phrase (Ford, 1990, pp 120-162). Type one consists of a single verb stem with pronominal prefixes and affixes indicating tense, aspect and modality, as in ka-djen-mene, 'it [the tide] is coming in' (chapter 7, track 7, text phrases 1-2). Type two consists of two verb stems, one of which inflects for person, number and gender by means of pronominal prefixes. For example, in ngala-viyitj nya-mu-nganggung 'sit and clap hands for us both' (chapter 7, track 8, vocal section 2, text phrase 2), nya-mu acts as a classifier to the preceding uninflected co-verb (-viyitj). Both types of verb phrase may begin with an incorporated body-part nominal (in the preceding example, this is ngala). The great majority of Batjamalh verb phrases are of type one; only ten classifying verbs have been identified for Batjamalh (Ford, 1990).
Batjamalh has two sets of pronominal prefixes. The first shows that it is transitive, requiring a Subject and Object. The second set shows that it is intransitive, requiring only one argument, the Subject. All other wangga languages have only one set of Subject prefixes that serve for both transitive and intransitive verbs and show the Object simply by suffixing one of a set of Object bound pronouns to the classifier verb stem.
Examples of Batjamalh intransitive prefixes are nga-, 'I,' as in nga-mi, 'I sit' (many examples throughout chapter 4 and chapter 7); and ka-, 'he,' (inflected for Realis mood) as in bandawarra-ngalgin ka-djen-mene, 'it [the tide] is coming in at Bandawarra-ngalgin' (chapter 7, track 7, text phrases 1-2). The pronominal prefix for 'you (one person)' inflected for Irrealis mood is nyV-, as in nye-bindja-ng, 'you sing' (chapter 4, track 7, vocal section 1, text phrase 2); nya-muy-ang, 'dance' (chapter 7, track 8, text phrase 2); nya-ngadja-barra-ngarrka, 'you (one person), come back here to me!' (chapter 4, track 24, text phrase 2).
Batjamalh has another set of portmanteau Subject/Agent + Object (S/A + O) pronominal prefixes for transitive verbs that show who is doing what to whom. The term portmanteau means that the forms are so fused and so old that it is impossible to unpack them to distinguish Subject from Object. For example, the pronominal prefix ngan- on transitive verbs can mean either 'he/she does it to me' or 'I do it to them.' Although it is impossible to unpack this portmanteau morpheme further, context will often disambiguate the meaning. For example ngan-dji-nyene (chapter 7, track 1, text phrase 6) could mean either 'I eat them' or 'he/she eats me.' In his spoken gloss, Lambudju is careful to disambiguate this by adding the serial verb nga-mi (I sit) to the phrase thus ngan-dji-nyene nga-mi in order to unambiguously produce the meaning 'I [sit and] eat them.' Nevertheless from the point of view of poetics it is interesting to reflect on why Lambudju choses to leave out the disambiguating element nga-mi in the sung form, particularly since in another related song 'Winmedje' (chapter 7, track 14) he retains nga-mi in the sung version of this phrase. Another instance from Lambudju's repertory where serial construction disambiguates a potential ambiguity with regard to ngan- may be seen in ngarn-rdut-mene ka-bara 'he has gone and left me behind' (chapter 7, track 10, text phrase 1b).
Similarly ambiguous is the pronominal prefix yV-, which could mean either 'you (one person) do it to him' of 'he/she does it to him.' Thus the phrase müng ya-mara (chapter 7, track 10, text phrase 3) could mean either 'you must catch him up!' or 'he/she must catch him up.' The presence of the serial verb nya-buring ('you do it deliberately') thus müng ya-mara nya-buring, not only adds aspectual information—that the action is deliberate—but it also disambiguates the meaning. Because the subjects of the two verbs have to be co-referential, the meaning must be 'you must [deliberately] catch him up.' Because this is rather a mouthful, in chapter 7 we translate this with the more idiomatic 'Catch him up!'
In the other wangga languages, direct or indirect object bound pronominals are suffixed to the classifier verb stem. In Batjamalh, only indirect object bound pronominals are suffixed to the verb stem. Examples of these are -ngarrka, 'for me,' (chapter 4, track 14, text phrase 4; chapter 7, track 17, text phrases 1-2); -nganggung,' for you and me,' (chapter 4, track 15, text phrase 5; chapter 7, track 17, text phrases 3-4) and ‑nüng 'for him' (chapter 7, track 13, text phrases 1-4).
Modality expresses the attitude of the speaker towards what s/he says. Batjamalh speakers vouch for the reality of their utterances by grammaticalising modality. Pronominal prefixes to the verb are marked Realis, if the speaker can vouch for the truth of what s/he says, or Irrealis, if s/he cannot.
To complicate matters, some pronominal prefixes and many verb stems do not alter to show modality. In one conjugation, however, verb stems are suffixed with ‑mene/‑nene/‑nyene to mark Realis (Ford, 1990, p 178). This is the function of ‑mene in ka-djen-mene ' it [the tide] is coming in [and I can vouch for this]' (chapter 7, track 7 text phrases 1-2). It is also the function of ‑mene in ngarn-rdut-mene 'he leaves me behind [and I can vouch for this]' (chapter 7, track 10, text phrase 1b), of ‑nyene in ngan-dji-nyene 'I eat them [and I can vouch for this]' (chapter 7, track 1, text phrase 6) and of ‑nene in nyen-ne-ne, 'they see you [and I can vouch for it]' (chapter 7, track 16).
If the speaker cannot vouch for the truth of what s/he says the utterance is marked Irrealis. Thus utterances about the future or commands (which are intrinsically not yet realised and therefore cannot be vouched for) are all marked as Irrealis.
Irrealis modality is signified by three mechanisms: by attaching Irrealis forms of the pronominal prefixes to verbs (if these forms exist), by inserting an Irrealis morpheme between the prefix and the verb stem or by changing the verb stems from a Realis to an Irrealis form (if these forms exist), sometimes by adding the verb-phrase-final purpositive enclitic ‑nung.
Commands combine an Irrealis prefix and a verb stem, which if such a form exists, will be inflected for Irrealis. For example, the utterance nye-bindja-ng nya-mu 'you [sit and] sing!' (chapter 4, track 7, vocal section 1, text phrase 2) is a command that combines two verb stems (sit and sing), both inflected for Irrealis modality. The pronominal prefix nye-/nya- (you) is in Irrealis form in both verbs. The verb stem ‑bindja (sing) does not change to show Irrealis modality but the serial verb does: ‑mu (sit) is an Irrealis form, whose Realis form is ‑mi (sit).
Statements of future intent combine all three mechanisms: a pronominal prefix inflected for Irrealis (if such a form exists), the Irrealis morpheme ‑p- and a verb stem inflected for Irrealis (if such a form exists). For instance, nga-p-pindja-ng nga-p-pur-ing-djü 'I'm going to climb up and go now' (chapter 4, track 22, text phrase 4), is a serial construction containing two verbs, the stems of which are ‑pindja- and ‑pur-. In both verbs the pronominal prefix nga- is one that remains the same irrespective of modality. Nga- is followed, in both verbs, by the Irrealis morpheme ‑p-. The verb stem bindja (here pronounced pindja to assimilate to the previous morpheme) is one that does not change to reflect modality. The second verb bur (here pronounced pur to assimilate to the previous morpheme) is, however, one that does change to reflect modality, in this case by adding the suffix ‑ing as well; its Realis form is bara.
Further to the use of Realis and Irrealis modalities in Batjamalh, we should note that there is an apparently anomalous construction that is unique to Batjamalh among the Daly languages, and which appears to contradict the normal semantic domains associated with Realis and Irrealis. This concerns prohibitions, which are negative commands. While, like all the Daly languages used in wangga, Batjamalh can express a prohibition by means of a clause-initial negator (in Batjalmalh this is nagulhü) and a verb inflected for Irrealis mood, Batjamalh has another productive mechanism for expressing prohibition. We see this in Barrtjap's song, 'Yagarra Delhi Nya-ngadja-barra-ngarrka' (chapter 4, track 24, text phrase 3): nanggang-gulhü kanya-bara-m, which consultants routinely translate as 'Don't be frightened!' Here the clause negator is realised as an enclitic particle ‑gulhü and the verb is inflected for Realis (rather than the expected Irrealis) modality.
The wangga texts show that some Batjamalh verb stems are inherently ambiguous. Despite its intransitive prefixes, the verb ‑bindja 'sing' functions as a transitive verb when Barrtjap and Lambudju add bangany 'song' to the verb phrase (e.g., chapter 4, track 1, track 4, track 10; chapter 7, track 2, text phrases 1-5; track 15, text phrases 1-4). Ka-bindja can, however, also mean 'he climbs up' or 'he hangs something up.' Only context, such as the evidence of the songman who composed the song, or the consensus of fluent speakers, decides that in Barrtjap's song, 'Ya[garra] Nga-bindja-ng Ngami' (chapter 4, track 7, vocal sections 2 and 3, text phrase 4) the serial construction nga-p-pindja-ng nga-p-puring-djü-nüng has the meaning 'I'm going to sing and then go back,' rather than that which it has in another of Barrtjap's songs ('Yagarra Tjüt Balk-nga-me Nga-mi'); namely, 'I'm going to climb up and go now' (nga-p-pindja-ng nga-p-puring-djü) (chapter 4, track 22, text phrase 4). In Song, Dreamings and Ghosts, Marett discusses in considerable detail instances where the songman exploits ambiguity, often to disguise deeper meanings (see for example Marett, 2005, pp 170, 171, 193-194).
Some verbs are restricted to gender-specific activities. For example, the verb stem ‑muya 'sway' is only used of women's dancing, so we can tell that the subject of the verb nya-muy-ang 'Dance!' must be female (chapter 7, track 8, vocal section 1, text phrase 2, track 17, text phrases 3 and 4). Similarly, the verb stem ‑mara, 'kick,' is only used of men's dancing, so the Subject of ya-mara 'Dance, man!' must be male (chapter 7, track 17, text phrases 3 and 4).
While serial constructions also occur in mundane texts and in conversation, they are especially common in Batjamalh wangga, amounting to 37.5% of all verbal clauses in Barrtjap's songs, and 29% of all verbal clauses in Lambudju's songs.
Serial constructions are conjoined clauses where a main verb is paired with an intransitive verb, which specifies whether the action denoted by the main verb it is habitual, repeated, intentional, permanent or temporary, and may add information about the physical orientation of the co-referential Subject. Both verbs are matched for tense, aspect and modality. For example, winmedje ngan-dji-nyene nga-mi means both 'I'm sitting eating oysters,' and 'I'm engaged in the temporary state of eating oysters' (chapter 7, track 14, text phrase 1). In this instance, the main verb is transitive, the serial is intransitive, but both are inflected for Realis modality. In nye-bindja-ng nya-mu 'You [sit and] sing!' (chapter 4, track 7, text phrases 2 and 4), both verbs are intransitive, and both are inflecMted for Irrealis modality.
In many Batjamalh serial constructions, the main verb is suffixed by the simultaneous marker ‑ng, glossed SIM, to show that the actions or states denoted by both main and serial verb occur at the same time. This marker does not occur in serial constructions in any other Daly language wangga.
Batjamalh mundane register contains at least 40 adverbial modifiers (Ford 1990:163-6). Batjamalh wangga texts contain just two, the modal adverb werret 'quickly' (chapter 4, track 24, text phrase 4; chapter 7, track 10, text phrase 3) and the temporal adverb yangarang 'today' (chapter 4, track 26, text phrase 4).
Of the fifteen free particles that occur in the Batjamalh mundane register (Ford 1990:166-69), ten are exclamations, two are negators, two interrogatives, and two are continuant particles.
Batjamalh wangga texts contain only three free particles: delhi 'Wait!' (chapter 4, track 24, text phrase 2); yagarra, literally 'oh no!' but untranslated in our song texts (chapter 4, tracks 1, 2, 3), with its allomorph yakerre (chapter 4, track 4, text phrase 3; chapter 7, track 28, text phrase 5); and karra (chapter 4, track 13, track 23; chapter 7, tracks 10, 11, 17, 23, 25), which always occurs at the start of a line and alerts the audience to the fact that what follows is sung by the dead (Marett, 2005, p 43). While karra is ubiquitous in the Marri Tjavin-Marri Ammu wangga, it occurs in only two of Barrtjap's songs and only five of Lambudju's songs. Because it occurs so frequently in the Marri Tjavin-Marri Ammu wangga, we do not in general include karra in our translations,
Fluent speakers of Batjamalh assured us that in the songs of Barrtjap the song word ya (e.g., chapter 4, tracks 1, 2, 3) is not the particle ya 'I don't know', which occurs in the mundane register. It is a vocable to fill out the rhythm, to provide a vehicle for extended vocalisations and to provide rhyme, as does the vocable ‑ngaya in bangany-nyung ngaya (chapter 4, track 4, text phrase 1).
Fourteen enclitic particles occur in the Batjamalh mundane register. The six most common of these enclitic particles also appear in wangga texts. They are:
the directional enclitic ‑barra 'towards speaker' (chapter 4, track 24, text phrase 2);
the continuant particle ‑djü, (chapter 4, track 7, vocal sections 2 and 3, text phrase 3);
the present enclitic ‑m (chapter 4, track 24, text phrase 3);
the perfective enclitic ‑maka (chapter 4, track 14, text phrase 2; chapter 7, tracks 10, text phrase 1b, 13, text phrases 1-4);
the purposive enclitic –nung (chapter 7, track 10), which Barrtjap habitually pronounces ‑nüng (chapter 4, tracks 7, 10, 15).
the temporal enclitic ‑bende  (chapter 7, tracks 10, 11, 12, 25) and its allomorph ‑nde 'now' (chapter 7, track 15);
the nostalgic enclitic ‑ve 'my eternal' (chapter 4, tracks 20  , 22; chapter 7, track 11, 23).
Although Barrtjap and Lambudju were both Wadjiginy songmen living and performing at Belyuen in the same time period, their wangga are very different. Marett (2005, p 92) points to Barrtjap's economy of lexicon, and contrasts it with Lambudju's texts which he characterises as 'highly varied in terms of both lexicon and structure' (Marett, 2005, p 182). Another way in which Lambudju differs strikingly from Barrtjap is in his use of two languages in a single text. He does this in track 11 of chapter 7, where he alternates text phrase in Emmi and Batjamalh. He is able to do this because the words make perfect sense to his multilingual audience. He begins with an expression that is the same in both languages karra-ve 'SW + forever', and continues with a serial construction in Emmi kanya-verver-rtedi kaya-ndhi, 'it [a breeze] is alway cooling my back'. The next text phrase begins with the same formulaic expression karra-ve, and continues with the Batjamalh phrase kak-ung-bende badjalarr, 'away now to Badjalarr'. He continues with a text phrase comprising the song words ribene ribene.  He concludes the song by reprising the opening line. For a possible explanation as to why he uses two languages in this song, see chapter 7, tracks 11 and 12.
We may conclude that Batjamalh wangga texts contain valuable evidence of the complex structures of this endangered language. The most recently composed songs also contain evidence that the language is moribund. For example, the text of 'Mubagandi' (chapter 7, tracks 25, 26), which was composed by Roger Yarrowin some time after the death of his brother-in-law Lambudju, contains a verb phrase yeme-ngadja, 'tell him to come back,' which fluent speakers of Batjamalh condemned as unacceptable in Batjamalh. Each of the component morphemes is perfectly good Batjamalh, but they are combined in an impossible way, in that the classifying verb precedes the intransitive co-verb. While this is common in Emmi, classifying verbs always follow their co-verb in Batjamalh. It looks as if Roger Yarrowin has constructed a Batjamalh verb that would be grammatical in Emmi, his mother-tongue. This is very different from the way Lambudju mixes languages. When Lambudju alternates text phrases in Emmi and Batjamalh they show no sign of mother-tongue interference. Rather the macaronic nature of the text reflects a period when there was a viable speech community bilingual in Batjamalh and Emmi, as evidenced by Lambudju's own history as a Batjamalh man who was brought up in an Emmiyangal family.
I analyse Emmi and Mendhe song language next. Emmi is the language of the Emmiyangal, whose traditional territory adjoined that of the Wadjiginy, and stretched south-west from the mouth of the Daly River to Mabulhuk (Cape Ford). Emmi has a closely-related dialect, Mendhe, the language of the Mendheyangal, whose territory adjoins Emmiyangal country, along the coast from Cape Ford to Cape Scott in the west and Nandhiwudi in the south. Inland of Mendhe country lie the Dashwood plains, which is Marrithiyel country. South of Nandhiwudi is Marri Ammu country.
Muluk (chapter 5) sang only in Mendhe, while Mandji (chapter 6) sang in Emmi, Mendhe and in Marri Tjavin (for example, chapter 6, track 8 is a mixture of Mendhe and Marri Tjavin).
As noted earlier, over at least the past one hundred and forty years, the Emmiyangal and Mendheyangal have, with their Wadjiginy affines, migrated from their traditional country to live in the Darwin region; specifically, on the coast of the Cox Peninsula, and since 1937, at the community of Delissaville, now known as Belyuen. There are Mendheyangal and Emmiyangal families at Belyuen, but very few fluent speakers of either Emmi or Mendhe. Due to the proximity of Darwin, only twenty minutes away by hydrofoil, at Belyuen there has been a gradual language shift to light Kriol, which is the first language of everyone aged forty or below, and the language of daily social intercourse between all Belyuen residents.
As also noted earlier, Emmi and Mendhe belong to the Western Daly linguistic subgroup, whose other members are Merranunggu, Marri Ammu and its closely related dialect Marri Tjavin, and Marri Ngarr and its dialect Magati Ke. This subgroup appears to go back several hundred years to a common mother language. Emmi and Mendhe have, however, diverged from these related languages because they have been influenced by, and have borrowed from, Batjamalh, the language of the Wadjiginy affines of Emmiyangal and Mendheyangal people. Geographical separation from Merranunggu, the closest linguistic relative to Emmi and Mendhe, has resulted in the loss of a phoneme and the initial segment of a pronominal. While Emmi and Mendhe share identical sound systems and grammar, they differ slightly in lexicon and morphology. Because the two language varieties are so close they are henceforth referred to as Emmi-Mendhe, except where relevant.
As regards phonemes, Emmi-Mendhe has lost Merranunggu's prestopped lateral, so Merranunggu pedle, 'white,' is pele in Emmi-Mendhe, and the Merranunggu bound pronominal -nedla, 'for his benefit,' is ‑nela in Emmi-Mendhe. As regards morphology, in current everyday Emmi-Mendhe, nginen means 'I sit', but Emmiyangal and Mendheyangal elders told me that anginen is the old-fashioned form. Anginen is transparently derived from ganginen, the Merranunggu form for 'I sit,' from which the Marri Ammu-Marri Tjavin and Marri Ngarr-Magati Ke form gangi is also derived.
Emmi and Mendhe share identical sound systems, shown in tables 3.3 and 3.4. These consonant systems differ from Batjamalh only in having a laminal dental series (stops and nasals), which is missing in Batjamalh. There is also a minor difference in vowel quality: where Batjamalh has a front rounded vowel /ü/, Emmi and Mendhe have a central unrounded vowel /ö/.
Emmi and Mendhe each have twenty-two consonant phonemes, including six full series of stops and nasals, which contrast for voicing. There are two apical series of stops and nasal (apico-alveolar and apico postalveolar or retroflex), two laminal series (lamino-dental and lamino-palatal), and two peripheral series (dorso-velar and bilabial). There are two phonemic fricatives: an apical /rz/ and a bilabial /v/. There are two laterals: one apico-alveolar, one retroflex and one lamino-dental. There are two rhotics: one apico-alveolar trill, and one retroflex continuant. There are two semi-vowels: one labial, one lamino-palatal. There are five phonemic vowels: high front /i/, low /a/, high back rounded /u/, open mid vowel /e/ and, unusually, central unrounded /ö/.
Table 3.3 Emmi-Mendhe consonant phonemes
Table 3.4 Emmi-Mendhe vowel phonemes
All these phonemes occur in Emmi-Mendhe wangga.
Differences between Emmi and Mendhe
While Emmi-Mendhe and Batjamalh have little vocabulary in common, sharing only 12% cognates in Dixon's 90-item Comparative Australian word list and 11% in the 400-item Comparative Australian word list, Emmi and Mendhe share 93% cognates from the 90-item word list and 96% cognates from the 400-item word list. Table 3.5 lists the lexical variants that exist in everyday Emmi and Mendhe. The bolded variants show up in wangga texts.
|ngala||ngula||'I use fingers'|
|ngarrinye-gurriny||tjanggurriny||'Belong to us, but not you (the addressee)'|
|ngula||ngala||'I chop it'|
|tjirrkinin||kunarra||'witchetty grub tree'|
|wurrum||wudut||'wet season; year'|
Table 3.5 Lexical variants between Emmi and Mendhe
Like Batjamalh, Emmi and Mendhe distinguish Realis from Irrealis modality. While they use the same basic mechanisms—changing pronominal prefixes, changing the main verb stem, inserting an irrealis morpheme and adding a verb phrase-final purpositive enclitic—the forms have little in common. 
Emmi and Mendhe, on the other hand, use virtually identical forms to mark Irrealis except that Mendhe omits the Irrealis morpheme. So the Emmi verb nga-wa-ni, 'I will walk,' is in Mendhe nga-ni. We see this in Muluk's famous Puliki song (chapter 5, track 1, text phrase 4): nga-ni-purr-mbele ngayi-nö alawa mari-pinindjela , 'I will always dance for you at Mari-pinindjela (Mica Beach)'.
Like Batjamalh, Emmi and Mendhe distinguish nominals (noun, adjective, pronoun, deictic), verbals, modifiers and particles. All these word classes are represented in the wangga song-texts.
Although, as in Batjamalh, a well-formed Emmi or Mendhe sentence need not contain an overt nominal, most Emmi-Mendhe wangga texts, like those in Batjamalh, present a significant proportion of overt nominals. Mandji's seventeen clauses contain twenty-two nominals, eight of which are incorporated into the verb. Muluk's seventy-two clauses contain one hundred and eighteen nominals, twenty-one of which are incorporated into the verb.
There is a wide range of nominals in these wangga texts, from the place names that locate the songs—duwun 'Duwun (Indian Island)' (chapter 6, track 1, text phrase 4); mari-pinindjela 'Mica Beach' (chapter 5, track 1, text phrase 4); and pumandjin 'Pumandjin Hill' (chapter 5, track 9, vocal section 1, text phrase 2)—to the common nouns alawa 'beach' (chapter 5, track 1, text phrase 4); piyamen.ga 'shady tree'(chapter 5, track 11, vocal section 3, text phrase 5); thawara 'mangrove sprout' (chapter 5, track 15, vocal section 1, text phrase 5); and dörr 'ground' (e.g. chapter 5, track 11). Some nouns, for example tjinbarambara 'seagull' (chapter 5, track 5, text phrase 1), and wak ‘crow’ (chapter 5, track 7, vocal section 1, text phrase 2) refer to Dreamings. Taken together these nominals reveal detailed knowledge of the totemic landscape known to songmen and their audience.
As in Batjamalh wangga, Emmi and Mendhe wangga use personal pronouns to draws the audience in and link them with the subject of the songs. The most common of these are: ngany 'I' (chapter 6, track 8, vocal section 1, text phrase 2), and the first person dual inclusive form nganggu-ga 'yours and mine' (chapter 5, track 7). The use of the anaphoric deictic yawa 'that place we know about' (e.g., chapter 5, track 7) heightens this intimacy.
Unlike the Batjamalh wangga, Emmi-Mendhe song texts contain no names of individuals, but like Batjamalh wangga, they do contain kin-terms, for example, mele 'elder brother' (chapter 6, track 8, vocal section 1, text phrase 2), which for an informed audience would normally be enough to identify a particular individual.
As in Batjamalh, Emmi-Mendhe nominals are marked for case and tense by ordered enclitics, for example:
- the dative case marker -nö, 'for,' and the temporal enclitic -endheni, 'now,' both of which appear in the following phrase from Mandji's wangga: mele ngany-endheni-nö, 'for my older brother now' (chapter 6, track 8, vocal section 1, text phrase 2);
- the ablative/causal marker -ngana, 'from,' and the perfective enclitic ‑yi, both of which appear in the following phrase from Muluk's wangga: wak-ngana-yi 'it came from crow' (chapter 5, track 7, vocal section 2, text phrase 1);
- the illocutionary enclitic ‑ndha 'really,' as in the expression yawa-ndha 'that place we really know about,' which occurs particularly frequently in Mendhe wangga texts.
Like Batjamalh, Emmi and Mendhe have a possessive pronoun, but the Emmi-Mendhe pronoun (unlike the Batjamalh pronoun) is marked for number and gender. Its base form is ‑gurriny 'belonging to.' This occurs in the following phrase from Mandji's wangga: ngandhi mandha na-gurriny, 'that song of his' (chapter 6, track 3, text phrase 3), where the possessive pronoun is marked by the pronominal na- to specify third singular male possessor. Another example is where Muluk signals his Mendhe identity by using the possessive pronoun to express his relationship to the shady tree that is the focus of his song: piyamen.ga ngani-gurriny, 'my shady tree' (chapter 5, track 11, vocal section 3, text phrase 5).
All languages used in wangga discussed in this book have another mechanism to express the possessive relation, whereby the possessed noun is juxtaposed with a free pronoun denoting the possessor, as in mele ngany 'my older brother' (chapter 6, track 8, vocal section 1, text phrase 2) and viye pumandjin, 'the top of Pumandjin Hill' (chapter 5, track 9, vocal section 1, text phrase 2).
Incorporated body-part nouns
As is common in everyday usage in all the languages used in wangga, Emmi-Mendhe wangga songs may incorporate body-part nominals into the verb phrase. Generally these are reduced forms of nouns denoting body parts. These incorporated body parts are sometimes used literally, as for example in ngammanya-mu-viye, ‘let’s both always keep dancing (with our hands above our heads)' (literally, let's both do it, do it to the head) (chapter 6, track 3, text phrase 3) and kana-nga-mu-viye karru, 'she [Numbali] is dancing, making a deliberate movement of her hands above her head' (literally, she's walking, doing it to her head as she goes) (chapter 5, track 9, vocal section 1, text phrase 2). Body-part nominals may also be used metaphorically, as when the idiom mörö-gumbu (literally 'buttock-foot') is used to mean 'from top to toe' or 'right through' in the serial construction nganya-bet-mörö-gumbu ngayi, 'let me always sing it for him right “through the night” [ie all night long]’ (chapter 6, track 8, vocal section 1, text phrase 2). Mandji also incorporates the non-body-part nominal mandha, 'song,' in gidji-djedjet-mandha-ya, 'he sings out that song' (chapter 6, track 1, text phrase 4).
Verbals and Verb phrase structure
In Ford (2007), I compared three Mendhe wangga texts with the prose versions of their subject matter, which were provided as explanation to Alice Moyle and recorded by her immediately after the wangga songs had been performed. In that paper, I showed that the spoken text reveal significantly less complexity with regard to verbal clauses than the wangga texts. These texts are reproduced in chapter 5 in the notes to tracks 1, 7 and 8.
In Batjamalh, Subject or Agent (S/A) and Object (O) are combined in a single portmanteau pronominal prefix to the verb. Emmi-Mendhe verbs are quite different in that the Subject is marked separately from the Object. Emmi-Mendhe verbs begin with a pronominal prefix that specifies the person, number and modality of the Subject or Agent (S/A) of the verb. This is followed by a verb stem, which in turn, is followed by a bound pronominal specifying the person, number and gender of the Direct Object (O) or Indirect Object (IO). There is no simple example of this construction in the Emmi-Mendhe wangga. An example from every speech might be, however: ka-ya-na-wut, 'he always gives it to him,' where the classifying verb stem is ‑ya 'lie down/always do' and the co-verb is ‑wut 'give.' The third person subject is marked by ka 'he,' the third person and the indirect object is marked by na 'to him.'
As in Batjamalh, the verb paradigms are irregular and cannot be fully predicted, and ordered verb-final enclitics signal aspect, tense, mood, modality, direction and degrees of illocutionary force. Unlike Batjamalh, however, the majority of Emmi-Mendhe verbs are inflected classifying verbs, which co-occur with an uninflected co-verb. The classifying verb describes the type of action/event or state specified by the co-verb. The complex verb phrases ka-na-kalkal-rtadi (chapter 5, track 7, vocal section 1, text phrase 1) and ka-na-putput-rtadi (chapter 5, track 7, vocal section 4, text phrase 1), both have ‑na 'go/walk' as the classifying verb, which here co-occurs with the co-verbs ‑kalkal 'climb' or ‑putput 'walk'. When used as a classifying verb, ‑na 'go/walk' also conveys aspectual information, specifying the action denoted by the co-verb as ongoing. Some classifying verbs, like ‑me 'say/do/feel' and ‑nya 'make/cook it,' may co-occur with any of dozens of co-verbs; other classifying verbs are attested as co-occurring only with a single co-verb. Twenty-seven of the thirty classifying verbs may also occur on their own, without a co-verb, as fully-inflected simple verbs.
Some simple verbs are more common than others, for example, the five intransitive verbs ‑nen 'sit'; ‑ma 'stand'; ‑ya 'lie down', ‑na 'walk' and ‑rru 'travel', which may also function as the second verb in serial verb constructions. Only two of these occur in Emmi-Mendhe wangga as the second verb in a serial construction. They are ‑ya 'lie down/always do,' and ‑rru 'travel/do deliberately.' ‑Ya occurs in thirty-five serial constructions, that is, in every Emmi or Mendhe wangga text except one, while ‑rru occurs three times, on each occasion with reference to a dance movement. ‑Ya is the verb of choice because it underscores the never-ending nature of the events celebrated in wangga.
Headless relative clauses
In spoken Emmi-Mendhe, as in spoken Batjamalh, headless relative clauses most often occur with a locative sense. This is true of some but not all relative clauses in the Emmi-Mendhe wangga texts. Consider, for example, Muluk's phrase nganggu-ga kaya yawa-ndha 'our X that is always in that place we know about.' Just what that X is, is clarified by the context. In one of Muluk's songs it is 'our seagull', but in another it is 'our stuff' (chapter 5, tracks 5 and 7). Other relative clauses provide information about the Subject of the verb, for example ka-me-ngana-yi ka-ya 'this [song] came from the one who always sings this' (chapter 6, track 7, text phrase 2).
Like Batjamalh, Emmi-Mendhe has a small closed class of words that modify the verb as temporal, locational or manner adverbs. None of these appear in the wangga texts of either Muluk or Mandji.
Like Batjamalh, Emmi-Mendhe has a small closed class of particles, which are of two types: 'free particles,' which occur as independent words, and 'enclitic particles,' which must attach to a host word, which can be from any word class.
While in general free particles are exclamations, like Puvuy! 'Keep going!' or tags such as ‑ening 'is it?' or wakkay 'it's finished,' they may also include polarity markers such as Yu 'yes!', or clause negators such as way. However, yakarre (chapter 5, track 9, vocal section 1, text phrase 2; chapter 6, track 3, text phrase 3) is the only free particle to appear in Emmi-Mendhe wangga.
Twenty-three ordered enclitic particles have been identified for the Emmi-Mendhe mundane register. Nine of these are adnominal enclitics, which attach to the end of a nominal phrase, where they generally mark case (see above under Nominals). The remaining fourteen are propositional enclitics, which attach to the end of a verb phrase, and modify the whole clause, providing information about direction, aspect, tense, modality, mood and illocutionary force. Adnominal enclitics precede propositional enclitics.
In Emmi-Mendhe wangga the only adnominal enclitic to occur is ngana 'from' as in ngany-ngana-yi 'it is from me' (chapter 5, track 11, vocal section 3, text phrase 4). Here, as in mundane usage, the adnominal enclitic ‑ngana precedes the propositional perfective enclitic yi.
Six of the fourteen possible propositional enclitics occur in the wangga corpus. These are:
- the tense markers ‑endheni 'now' (chapter 6, track 8, vocal section 1, text phrase 2);
- the perfective marker ‑(e)yi (chapter 5, track 7, vocal section 2, text phrase 1; chapter 6, track 1, text phrase 4);
- the purposive marker ‑(e)nö (chapter 5, track 1, vocal section 1, text phrase 2; chapter 6, track 8, vocal section 1, text phrase 2);
- the topic marker ‑ga (chapter 5, track 5, text phrase 2);
- the illocutionary marker ‑ndha (chapter 5, track 2, vocal section 1, text phrase 2);
- the directional marker ‑ya 'thither' (chapter 6, track 1, text phrase 4).
Marri Ammu and Marri Tjavin
The Marri Ammu people have as their ancestral country the coastal strip south from Ngandhiwudi to Anggileni (Cape Dombey). Their territory abuts that of the Mendheyangal to the north, the Marrithiyel people to the east, and the Marri Tjavin people to the south. Marri Tjavin territory stretches from Nadirri at the mouth of the Moyle River, inland to Perrederr, Yendili and Yenmilhi. It is bounded on the east by Marri Ngarr country and on the south by Magati Ke country. Today, almost all Marri Ammu and Marri Tjavin families live at Wadeye, but they visit their country regularly when, as in the dry season, the weather and finances permit.
The Walakandha wangga (chapter 8) has been composed by a number of Marri Tjavin songmen over the past four decades or so. It is the main song series for many of the Marri language groups resident at Wadeye, including the Marri Ammu who participate in ceremonies primarily as dancers. The Marri Ammu have their own series, the Ma-yawa wangga (chapter 9), but this is no longer performed today, and has probably not been used in ceremony for several decades (Marett, 2005, p 135).
Marri Ammu and Marri Tjevin are very closely related dialects, differentiated by only a few lexical items. Their closest linguistic relative is Marrithiyel, which has independently undergone some minor sound changes and diverged in its lexicon. Less closely related is Marri Ngarr and its dialect Magati Ke, and related more distantly still are Merranunggu and Emmi-Mendhe. All these language varieties have been identified by Ian Green as belonging to the Western Daly linguistic subgroup (Green, 1994).
Members of the Marri Ammu and Marri Tjavin language groups differentiate themselves from these other groups by calling themselves Ma Thanggural. Marri Ammu wangga are known as wangga ma yawa 'wangga belonging to the ma yawa'. Marri Tjavin wangga are known as wangga ma walakandha 'wangga belonging to the ma walakandha.' Ma yawa and ma walakandha are the spirits who gave the songs to the songmen (Marett, 2005).
Marri Tjavin and Marri Ammu are so similar in phonology, morphology and syntax that they can be discussed as one system. I will, however, point out the small differences that do exist where relevant. I will show what we can learn of this system first from Walakandha wangga texts,and then from Ma-yawa wangga texts.
Marri Ammu and Marri Tjavin share identical sound systems, as shown in tables 3.6 and 3.7. This differs from the Emmi-Mendhe sound system only in having an extra fricative, the lamino-palatal /sj/ and in lacking Emmi-Mendhe's fifth vowel, the unrounded /ö/. They each have twenty-eight consonant phonemes, including six full series of stops and nasals, which contrast for voicing. There are two apical series of stops and nasal (apico-alveolar and apico postalveolar or retroflex), two laminal series (lamino-dental and lamino-palatal, and two peripheral series (dorso-velar and bilabial). There are three phonemic fricatives: an apical /rz/, a laminal /sj/ and a bilabial /v/. There are three laterals: one apico-alveolar, one apico-postalveolar and one lamino-dental. There are two rhotics: one apico-alveolar trill, and one retroflex continuant. There are two semi-vowels: one labial, one lamino-palatal. There are four phonemic vowels: high front /i/, low /a/, high back rounded /u/ and the open mid vowel /e/. Before a following high front vowel /i/, the velar nasal /ng/ becomes palatalised, and sounds like ngyi, as happens in the neighbouring language, Murriny Patha.
All these phonemes occur in Marri Tjavin and Marri Ammu wangga.
Table 3.6 Marri Tjavin-Marri Ammu consonant phonemes
Table 3.7 Marri Tjavin-Marri Ammu vowel phonemes
Marri Tjavin and Marri Ammu share 98% cognates from Dixon's 90-item Comparative Australian word list and 98% cognates from Dixon's 400-item comparative Australian word list. Table 3.8 lists the lexical variants between mundane Marri Ammu and Marri Tjavin. The bolded italicised variants show up in wangga texts.
According to Tryon (1974, p xiv), Marri Tjavin-Marri Ammu share approximately 43% cognates with Emmi-Mendhe and approximately 18% cognates with Batjamalh. However, Tryon's early work was based on a limited sample. No later comparative data exists.
|Marri Tjavin||Marri Ammu||English gloss|
|-nginanga||-nginyanga||'against my will'|
Table 3.8 Lexical variants between Marri Ammu and Marri Tjavin (bolded forms occur in wangga)
Where forms peculiar to Marri Ammu occur in wangga, as they do in 'Yendili No. 5' (chapter 8, track 29), they serve to emphasise the Marri Ammu identity of the speaker, in this instance, Honorata Ngenawurda, Frank Dumoo's mother.
Marri Tjavin and Marri Ammu distinguish the same word classes as Batjamalh and Emmi-Mendhe. These are all represented in Walakandha wangga and Mayawa wangga.
As in Batjamalh and Emmi-Mendhe wangga, there are more overt nominals in Walakandha wangga than one would normally expect in a mundane text (see 4.5 below). For instance, the earliest Walakandha wangga songs (composed by Stan Mullumbuk), comprise five clauses containing fourteen nominals. Seven of these (50%) are nouns that introduce the Subject of the clause, two are body-part nouns incorporated into the verb phrase, four are locative noun phrases incorporating a body-part noun, and one is a deictic marked with ‑gu as a different topic.
The range of nominals is similar to that seen in wangga already analysed. As in other wangga repertories, proper nouns frequently occur in Walakandha wangga, whether they be names of people—for example Wutjelli (chapter 8, tracks 3, 22); Berrida, Munggumurri (chapter 8, track 35); Tjagawala (chapter 8, track 27)—or of places—for example Yendili (chapter 8, tracks 10, 13, 14, 29, 30, 32, 35); Yenmilhi (chapter 8, tracks 11, 20); Truwu (chapter 8, tracks 16-18, 22); Kubuwemi (chapter 8, tracks 12, 33); Nadirri (chapter 8, track 19); Yenmungirini (chapter 8, track 34); Pelhi (chapter 8, track 20); Pumurriyi (chapter 8, track 24); Lhambumen (chapter 8, track 31); Ngumali (chapter 8, track 36); Kinyirr (chapter 8, track 37); Namadjawalh, Yimurdigi, Kanbirrin, Lhambudinbu (chapter 8, track 38); Rtadi-wunbirri (chapter 8, track 39). A frequently used non-specific reference to place is the important phrase nidin-ngin, which has been various translated as 'poor fellow my country,' 'poor bugger my country,' and 'my dear country.' It is a powerfully emotive articulation of people's attachment to country.
Some common nouns provide situational context, for example mirrwana 'cycad palm' (chapter 8, track 21), which is traditionally associated with Walakandha, and ngatha devin bugim rtadi 'a solitary house with a white roof' (chapter 8, track 23), which refers to a particular house at the Nadirri outstation.
Kin-terms also occur in Walakandha wangga, including mana 'older brother' (chapter 8, track 20, text phrase 1); angga 'grandfather/grandson' (chapter 8, track 27, text phrase 3), thidha 'father' (chapter 8, track 25, text phrase 1), ngatja 'child' (in Marri Ammu) (chapter 8, track 29, text phrase 3). The composers of Walakandha wangga use these to draw the audience in, just as Barrtjap and Lambudju did in Batjamalh wangga. Like Barrtjap, Muluk and Mandji (discussed above), they also create intimacy by using cardinal pronouns, for example nany (chapter 8, track 25, text phrase 1), yigin 'I' (chapter 8, track 25, text phrase 2), ngangga-nim 'for all of us' (chapter 8, track 20, text phrase 4).
The densest use of nominals occurs in the songs of the late Marlip Philip Mullumbuk. His fifty-six clauses contain sixty-nine nominals, an average of 1.25 nominals per clause. Most of these are place names, and thirty-nine occurrences of the place name Namadjawalh come from a single wangga text 'Wedjiwurang' (chapter 8, track 38), about wedjiwurang 'wallaroo', the totem whose dreaming-site is Namadjawalh. Mullumbuk's use of the anaphoric locative yivi-ndja (chapter 8, track 38, vocal section 1, text phrase 2) is cognate with Muluk's formulaic noun phrase yawa-ndha 'that place we know about.' 
Like Batjamalh and Emmi-Mendhe, Marri Tjavin-Marri Ammu has a system of noun classifiers, but its system is much more comprehensive and exact. Its twelve noun classifiers differentiate male from female, edible vegetables from meat, edible vegetables from grasses, fire, danger, liquids, language, spears, inanimate objects, and location in time or space. Each of these categories of nominal has its own classifier. When first introduced into mundane discourse, a species item is always preceded by an appropriate noun classifier. Thereafter, the referent may occur without the classifier, or be referred to by the noun classifier without the species name. For instance, walakandha 'ancestral dead' would in mundane register be first introduced into the text with the male human classifier ma, as ma walakandha. In Walakandha wangga texts, however, it never occurs with a noun classifier. Indeed, except for Marlip Philip Mullumbuk's anomalous  song 'Wedjiwurang' (chapter 8, track 38), there are no noun classifiers in the Walakandha wangga texts. In this song Marlip introduces the topic with the generic animal classifier awu without the species name; only in the next text phrase does he name the species as the topic wedjiwurang-ga, marking it with the topic marker ‑ga (chapter 8, track 38, vocal section 1, text phrases 1 and 2).
Complex noun phrases
Unlike Emmi-Mendhe, Marri Tjavin-Marri Ammu has only one way to express the possessive relation. This is by juxtaposing possessed and possessor nominals, as in thidha ngany 'your father' (literally 'you father') (chapter 8, track 25, text phrase 1) and yeri meri yigin 'my children' (literally 'children males I' (chapter 8, track 34, text phrase 2). Marzi mungirini 'deep inside the jungle' (chapter 8, track 2, text phrase 2) is a possessive noun phrase where the possessed entity is a body part used metaphorically to represent a geographical feature shaped like the body part.
Verb phrase structure
Like Batjamalh and Emmi-Mendhe, Marri Tjavin-Marri Ammu verb phrases consist of a series of ordered morphemes. Marri Tjavin-Marri Ammu and Emmi-Mendhe verb phrases have identical structures, in that they begin with a pronominal prefix that specifies the person, number and modality of the Subject/Agent of the verb.
Then comes the main verb stem, which is followed by a slot for the Object of the verb. If there is a co-verb, this will come next. Then comes a slot for an incorporated noun, followed by a slot for a benefactive bound pronominal, followed by a slot for an adversative bound pronominal. This is followed by a series of ordered slots for enclitic particles, which specify number and gender of Subject and/or Object, tense and aspect.
In Walakandha wangga, a noun incorporated into the verb phrase conveys more than just the literal meaning, for example, venggi 'knee' when incorporated into the verb-phrase in ki-nyi-ni-venggi-tit 'he bent his leg', means one knee bent over the other, in the position known in Aboriginal English as 'number four leg'. Similarly, the noun mi 'eye' incorporated into the verb phrase in 'Yendili No. 5' (wudi yendili ngil-dim-mi-nginyanga  -ndjen 'I'm closing down the spring at Yendili against my will'; see chapter 8, track 29, text phrase 4) means the round, eye-shaped heart of the water-hole.
Unlike Emmi-Mendhe wangga, Walakandha wangga contain no benefactive bound pronominals; on the other hand, unlike all other wangga, a significant proportion of Walakandha wangga contain adversative bound pronominals. The most commonly occurring form is ‑nginyanga 'I couldn't stop it' (chapter 8, track 29, text phrase 4).
Like Emmi-Mendhe, Marri Tjavin-Marri Ammu has thirty inflected verbs, but their forms are only distantly related. In both languages the verb paradigms are irregular and cannot be fully predicted. Many of the Marri Tjavin-Marri Ammu forms have neutralised. Twenty-seven of these inflected verbs may co-occur with a following uninflected co-verb.
As in Emmi-Mendhe, the classifying verb describes the type of action/event or state specified by the co-verb. For instance, the intransitive verb ‑wa 'stand' classifies co-verbs which involve vertical action. Many collocations are transparent, for example purangang kuwa-vapa 'the waves are crashing' (literally, 'the salt water stands up and crashes') (chapter 8, track 16, text phrase 1).
In addition to their primary meaning, classifying verbs may also provide aspectual information. So, for example, the primary meaning of the intransitive verb ka-ni  is 'go/walk.' The first person Irrealis form of this is ngumbu and this is what it means in ngumbu-vup-nim 'let's all get going now' (chapter 8, track 26, text phrases 1-2). But ka-ni 'go/walk' also renders an action ongoing, so ka-ni-wurr-a means 'he was dying' (chapter 8, track 38, vocal section 4, text phrase 5). Likewise, the transitive classifying verb ka-rri 'use hands' co-occurs with co-verbs denoting actions done with the hands. Ka-rri-wuwu 'it flies' (chapter 8, track 39, text phrase 2) is a transparent collocation meaning that the bird is using its wings (hands) to fly. But ka-rri is also used of an external agency forcing an intransitive action, so ka-rri-tik 'the tide goes out' (chapter 8, track 19, text phrase 1), which is a metaphor for dying, means an external force makes the tide go out or the person die. Likewise again, the primary sense of the transitive classifying verb ki-nyi is to 'make something', but this verb is also a causative verb, meaning 'make something happen.' So ki-nyi-ng-kurr-a means literally 'it [a breaker] made me get hit' (chapter 8, track 27, text phrases 1-2). Ki-nyi also transitivises intransitive verbs. So, for example ki-nyi-ni venggi-tit means 'he makes himself bend his knee' (chapter 8, track 2, text phrase 2).
Some classifying verbs, like ki-nyi 'make something', ka-rri 'use hands'; kin 'paint, draw' and the zero stem inflected verb ki-Ø, may co-occur with any of several co-verbs, but there are many more co-occurrence restrictions in Marri Tjavin-Marri Ammu than in Emmi-Mendhe, and some classifying verbs are attested as co-occurring only with a single co-verb.
The Walakandha wangga texts contain a small subset of inflected verbs, twelve out of a possible thirty. They are: ku-rzi 'sit'; ku-wa 'stand'; ka-vulh 'lie down'; ka-ni 'go/walk'; ka-yirr 'go/travel'; ki-l 'chop'; ki-n 'paint, draw'; ki-nyi 'make something'; ka-rri 'use hands'; ki-din 'see'; ku-munit 'pick up'; and ku-muyi 'emerge'. Ten of these inflected verbs function as classifying verbs to one or more co-verbs in the wangga texts. Ka-vulh 'lie down'; ku-munit 'pick up'; and ku-muyi 'emerge' occur only as simple verbs.
Simple verbs as Serial verbs
In mundane usage, the five intransitive verbs—ku-rzi 'sit', ku-wa 'stand', ka-vulh 'lie down', ka-ni 'go/walk' and ka-yirr 'go/travel'—also function as the second verb in serial verb constructions. Only three of these occur as the second verb in wangga serial constructions. These verbs are ka-vulh 'lie down/always happen', which occurs in twenty-one serial constructions, ka-ni 'go/walk; keep on doing', which occurs in eight serial constructions; and ku-wa 'stand', which occurs in four serial constructions.
The second verb in a serial construction provides aspectual information about the action denoted in the co-verb, and also specifies the physical orientation of the co-referential Subject of the two conjoined clauses. For example, where ka-ni is the second verb, the repetitive nature of the action expressed in the co-verb is emphasised, where ka-vulh 'lie down/do always' is used, the permanence of the action is stressed. When ku-wa is used, only vertical orientation is stressed.
Sometimes this physical orientation is incompatible with the meaning of the co-verb and then the aspectual information overrides the literal meaning. Sometimes it is merely ambiguous whether the literal or aspectual meaning should prevail and this ambiguity is frequently exploited in wangga texts.
Walakandha songmen, like the Ma-yawa wangga songmen to be discussed below, frequently emphasise the everlasting quality of the actions they describe by using ka-vulh 'lie down/do always' as the second verb in most of these serial constructions.
Headless relative clauses
The only Walakandha wangga composer to use headless relative clauses is Marlip Philip Mullumbuk. This is a locative relative clause that provides information about Yenmungirini: kangi-nginanga yenmungirini na pumut-pumut kurzi, 'I've got to stay here at Yenmungirini where the Headache Dreaming is' (chapter 8, track 34, text phrase 2).
Like Emmi-Mendhe, Marri Ammu-Marri Tjavin has a small closed class of words that modify the verb as temporal, locational or manner adverbs. The only ones to occur in the Walakandha wangga are wandhi 'behind' (chapter 8, track 20, text phrase 3; track 34, text phrase 1) and warambu 'high up' (chapter 8, track 36, text phrase 1). In track 20, wandhi occurs outside the verb phrase. In track 34, wandhi is incorporated into the verb phrase, as is warambu in track 36. The absence of other independent modifiers is accounted for by the incorporation within noun phrases and verb phrases of spatial nouns such as ‑rtadi 'back/top surface' and marzi 'belly/inside'.
Like Emmi-Mendhe, Marri Ammu-Marri Tjavin has a small closed class of particles, which are of two types: 'free particles,' which occur as independent words, and 'enclitic particles,' which must attach to a host word that can be from any word class. Few free particles occur in Walakandha wangga. They are: the negator ambi 'not' (chapter 8, track 25, text phrase 1); the exclamations wakkay 'it's finished!' (chapter 8, track 27, text phrase 3) and yakerre (chapter 8, track 36, text phrase 3); and the song word karra, which begins many grammatical text phrases in Walakandha wangga and is absent in only three songs.
Ordered enclitic particles occur in all Walakandha wangga . They are all propositional enclitics, which attach to the end of a verb phrase, and modify the whole clause, providing information about direction, aspect, tense, modality, mood and illocutionary force. One propositional enclitic, the directional -wurri 'to me', occurs in half (nineteen) of the thirty-eight wangga songs in chapter 8. The other enclitics which occur in Walakandha wangga are:
- the perfective marker -(ey)a 'it is over' (chapter 8, track 1, text phrase 1);
- the tense marker-ndjen 'now/then' (chapter 8, track 26, text phrase 1);
- the purposive enclitic -(e)ni (chapter 8, track 32, text phrase 1-2);
- the illocutionary marker –(a)ndja (chapter 8, track 33, text phrase 1);
- the ablative/causal enclitic -nganan, which occurs in Marlip Philip Mullumbuk's 'Walakandha No. 5': kurriny-rtadi-warambu-nganan-wurri-ya 'they came towards me from high in the inland country' (chapter 8, track 36, text phrase 1);
- the allative enclitic -rzan (chapter 8, track 25, text phrase 2);
- the reiterative enclitic 'again' -da (chapter 8, track 38, vocal section 1, text phrase 3).
The topic marker -ga occurs only in the wangga of Marlip Philip Mullumbuk, where it modifies a noun (chapter 8, track 33, text phrase 1). Different topic -gu also modifies a nominal (chapter 8, track 33, text phrase 1), but is used propositionally elsewhere (chapter 8, track 7, text phrase 1). Marlip Philip Mullumbuk is the only composer to contrast -ga and -gu (chapter 8, track 33, text phrase 1).
A conspicuous feature of the Walakandha wangga texts is that they do not contain the song words that play such an integral role in Batjamalh and Emmi-Mendhe wangga. The only song word used in the Walakandha wangga is karra, which is used to signal that the following statement has emanated from the ancestral realm. In one instance (chapter 8, track 28), the text contains no words other than karra.
I will conclude my discussion of Walakandha wangga by noting three unusual texts. The first text 'Yendili No. 2' (chapter 8, track 14) breaks with convention, in that it was composed in Marri Ngarr  by a Marri Ngarr woman, Maudie Attaying Dumoo, about a Walakandha Dreaming site in Marri Tjavin country for which her Marri Tjavin husband had responsibility. The song is addressed to the couple's children, who are bilingual in Marri Ngarr and Marri Tjavin and inheritors of responsibility for the site. It addresses the children directly, and orders them to keep hold of the place named in the song. For more esoteric meanings, see the notes to track 14 in chapter 8.
The first two text phrases are identical and consist of a pair of complex verb phrases inflected for Irrealis modality; the final clause is a verbless vocative phrase: aa ye-ngin-a 'Ah, my dear children.' The first two text phrases begin with karra followed by the reduplicated place name 'Yendili.' Even though the text is in Marri Ngarr rather than Marri Tjavin, its structure follows that of other Walakandha wangga (Marett, 2005, pp 125-126).
The second unusual wangga text, Mandji's 'Karra Mele Ngany-endheni-nö' combines features of Mendhe wangga and Marri Tjavin wangga (chapter 6, track 8). Here Mandji alternates vocal sections in Mendhe and Marri Tjavin. The Marri Tjavin vocal section includes an adversative bound pronominal ‑ngangga, the directional enclitic ‑wurri, a serial construction and a headless relative clause, all of which also occur in the Walakandha wangga.
A third anomalous text, Philip Mullumbuk's 'Wedjiwurang' has been discussed in detail elsewhere (see Ford 2007 and notes to track 38 in chapter 8).
Differences between the Ma yawa wangga and the Walakandha wangga
The twenty-nine Ma-yawa wangga tracks comprise twelve songs. I exclude from my analysis 'Walakandha ngindji' (chapter 9, tracks 1 to 3), because it is borrowed wholesale from a Walakandha wangga text (chapter 8, track 23). Nor do I take into account the performance of 'Thalhi-ngatjpirr' sung by Frank Dumoo and Colin Worumbu Ferguson (chapter 9, track 24), in which a Marri Tjavin songman borrows wholesale a Ma-yawa wangga text.
The remaining texts reveal the following differences from the Walakandha wangga texts.
The texts contain 224 nominals in 146 clauses, approximately 1.5 overt nominals per clause, much higher than in mundane usage, but roughly comparable in frequency to the nominals in Walakandha wangga texts.
These texts contain fewer place names than the Walakandha wangga, but the occurrence in them of geographical terms such as mungirini 'jungle' (chapter 9, track 16, text phrase 2), diyerr 'cliff' (chapter 9, track 20, vocal section 1, text phrase 1), wudi-pumininy 'freshwater spring' (chapter 9, track 21, vocal section 1, text phrase 1), and purangang 'saltwater' (chapter 9, track 20, vocal section 3, text phrase 1) fixes the location of the songs, as does the anaphoric locative kagan-dja 'that place we know about' (chapter 9, track 13, vocal section 1, text phrase 2).
As in other wangga repertories, intimacy is conveyed by the use of kin-terms, but these texts contain only the kin-term mana 'elder brother', which occurs 21 times, almost always introduced by the song word karra. In all these cases mana refers to Dreaming beings, which are commonly referred to as 'elder brother' throughout the Daly. As Marett has noted (2005, p 135), the Ma-yawa wangga focuses on Dreamings (ngirrwat) and Dreaming places (kigatiya) to a much greater extent than other wangga repertories. Of the seven place names that occur in the texts, one, menggani 'butterfly Dreaming' is also the name of a Dreaming place; in all, six Dreamings—Menggani, Malhimanyirr, Tjerri, Tjiwilirr, Tulh, Ma-yawa—are named. In one text, 'Malhimanyirr,' the Dreaming is explicitly described as kanyi-ngin 'my totem' (chapter 9, track 17, vocal section 1, text phrase 1).
Unlike the Walakandha wangga, these texts contain a range of generic nouns that function as classifiers. These are: ma 'male human' (chapter 9, track 20, vocal section 1, text phrase 2), wulumen 'old man' (chapter 9, track 4, vocal section 1, text phrase 1, track 28), miyi 'edible vegetable' (chapter 9, track 28, vocal section 1, text phrase 3), nidin 'place' (chapter 9, track 25, vocal section 1, text phrase 1), thawurr 'inanimate' (chapter 9, track 27, vocal section 1, text phrase 1).
Complex noun phrases
Like the Walakandha wangga, these texts contain adjectival noun phrases: meri ngalvu 'many people' (chapter 9, track 22, vocal section 1, text phrase 1), mungirini kapil 'big jungle' (chapter 9, track 16, vocal section 1, text phrase 2), mungarri kapil 'deep sleep' (chapter 9, track 18, vocal section 1, text phrase 4) and possessive noun phrases: mana nganggi 'our brother' (chapter 9, track 20, vocal section 3, text phrase 2).
Out of a total of 146 clauses, only 12 are verbless. Seventy-seven are simple clauses containing only one verb, and 59 are complex clauses that combine two verbs in a serial construction. The verb ka-vulh, 'he lies/does it forever,' occurs as the serial verb in 45 of these constructions; ka-yirr, 'he travels/does deliberately,' occurs in four; ka-ni, 'he goes/keeps on doing,' in four and ku-wa, 'he stands up,' and its Irrealis form ngunda in six. There are only four embedded, locative, headless relative clauses and these all express locative information. For example:
kani-djet diyerr kuwa, ' he sits where the cliff stands up [i.e., at the foot of the cliff]' (chapter 9, track 20, vocal section 1, text phrase 1);
kani-djet kuwa kagan-dja 'he is sitting right here where it [the cliff] stands up' (chapter 9, track 20, vocal section 2, text phrase 1);
nidin na kaddi devin kurzi 'country that is just for us' (chapter 9, track 22, vocal section 1, text phrase 1).
The texts contain a limited range of adverbs: nal 'just' (chapter 9, track 28, vocal section 2, text phrase 3); kisji 'just like this' (chapter 9, track 28, vocal section 1, text phrase 4) and wandhi 'behind' (chapter 9, track 15, vocal section 1, text phrase 2).
Ma-yawa wangga texts use only two enclitic particles: ‑dja 'really' (chapter 9, track 12, vocal section 1, text phrase 3); and ‑gu 'different topic' (chapter 9, track 22, text phrase 2). But -gu is not used in Ma yawa wangga as it would be in mundane usage, where -gu only occurs to mark a switch of topic, to contrast with an earlier topic, marked with ‑ga. In Ma-yawa wangga, the distinction between Topic marker -ga and Different Topic marker -gu is neutralised; ‑ gu marks any topic, occurring frequently—in 46 clauses—but never in contrast with a preceding ‑ga.
This is also true of early Walakandha wangga texts, for example, in Stan Mullumbuk's Track 4 text phrase 1, and track 7 text phrase 1), but not of the most recently composed Walakandha wangga, where Philip Mullumbuk always conforms to mundane usage. In his Track 34, verse 1, text phrase 1, the sole topic of the text phrase is marked by -ga, but in track 33, text phrase 1, and track 38 vocal section 3, text phrase 3, and vocal section 4, text phrases 4-5, the first topic is marked by -ga, and the second by -gu.
It is also true that in Ma-yawa wangga and Stan Mullumbuk's Walakandha wangga, -gu always occurs at the end of a text phrase, so, in early Walakandha wangga, -gu appears to function primarily as a metric filler. Only in later Walakandha wangga does its use conform with mundane usage. Maybe Philip Mullumbuk alludes to this by ending the final text phrase of the final vocal phrase of Track 38 with -gu.
In Ma-yawa wangga, -gu appears to be used both as neutralised topic marker and as metric filler. It occurs 46 times, in most instances (66%), at the end of a text phrase. But in 14 instances (30% of the total), it marks the noun phrase that begins the text phrase, and in one instance it marks the verb phrase at the beginning of the text phrase, in other words, the topic. Modern mundane usage would omit it, but Ngulkur plays with poetic and mundane usages, like a signature.
The Ma-yawa wangga contain a significant number of simple and reduplicated forms, ranging from adverbial wandhi-wandhi 'behind' (chapter 9, track 14, vocal section 1, text phrase 2) and kisji-gisji 'like this' (chapter 9, track 6, vocal section 1, text phrase 4), to the verbal kimi-gimi 'he does/says/sings' (chapter 9, track 10, vocal section 1, text phrase 1) and the partial reduplication ga-kap 'call out' (chapter 9, track 16, vocal section 1, text phrase 1). This use of reduplication extends to place-names, for example wudi-pumininy-pumininy (chapter 9, track 20, vocal section 2, text phrase 2) and rtadi-wunbirri-wunbirri (chapter 9, track 6, vocal section 1, text phrase 3).