[This document can be acquired from a sub-directory coombspapers via anonymous FTP and/or COOMBSQUEST gopher on the node COOMBS.ANU.EDU.AU] The document's ftp filename and the full directory path are given in the coombspapers top level INDEX file] [This version: 4 August 1993] -------------------------------------------------------------------------- File 2. There are 9 parts of this thesis: from 0 to 8 -------------------------------------------------------------------------- A Grammar of Garadjari, Western Australia by Anna Kristina Sands 1989 Bachelor of Arts Thesis Department of Linguistics, Faculty of Arts Australian National University, Canberra ACT 0200 2 PHONOLOGY 2.1 Phonemes 2.1.1 Consonants The phoneme system of Garadjari follows along the lines of a typical Australian language. There are five places of articulation in the consonant system with only one laminal series. Following is a chart of the consonant phonemes : Chart 2.1 Garadjari Consonantal Phonemes apical laminal bilabial alveolar post-alv. palatal dorsovelar stop b d rd dj g nasal m n rn nj ng lateral l rl lj rhotic r,rr(?) semivowel y w The rhotics present a problem because in Laves' texts he seems to distinguish at least two rhotics but not with great consistency, ie. he in many words he uses any form of the rhotic while in others he is very consistent. His method, in the earlier texts, of distinguishing the rhotics is to place a dot under some, mostly those followed by a consonant,but not exclusively ( ie. r ) and a curve under others ( ie. r ) while others yet he leaves plain. In rapid transcriptions it becomes difficult to distinguish between those two which he marks. Part way through his texts he changes his transcription to a plain r and an italicised r , however, he still includes in many places an r with a mark underneath and thus we get up to four different signs for a rhotic in the one text. We also find that there is no real consistency between the first two signs he uses and the second two ( for an example of this see the appendix ). It is thus very difficult to determine exactly what is happening and how many rhotics the language actually has. For the sake of simplicity I will use a single symbol, r, to cover the rhotics wherever they occur in the texts, but will assume that there are two rhotic phonemes to bring it in line with the majority of Australian languages. We can assume that these are a continuant, or approximant, and a trill though there is no indication in Laves' texts that this is so. A second area of difficulty is within the apicals, ie. do we have one or two series here ? We note that many words alternate quite freely between the alveolar and post-alveolar apicals in Laves' texts and we regularly get alternations such as : garli gali 'to dig' birdi bidi 'hole' From this we might like to suggest that there is only one apical series and that the above alternations are in free variation. However, we get other morphemes that never show such alternation. ie. the suffix -rnu for the past never appears as -nu despite its very common occurrence in the texts. To resolve the issue we would like to find some minimal pairs involving the two phonemes, and we do, indeed, have minimal pairs but not very many and these are all unclear, ie. I have found only the following : a) yada 'sheild' yarda 'kill' b) wanda- 'to stay' warnda ? ( one example only) c) guli 'fight' gurlu 'disabled' d) djirdu-ngka 'secretely' djidur 'all cut up' Now in a) as well as a difference in the apical we have the difference between a noun and a verb and these can generally not be confused because they take different inflections. In b) the second example is unclear and in c) and d) we only have near-minimal pairs. Thus, the evidence for two apical series is shaky at best, and the only real evidence we have that the sequence [ rn] etc. in Laves' transcriptions is not actually a sequence of rhotic plus nasal is that in his English translations of names that contain this sequence he transcribes this as n, ie. the traditional symbol used with retroflexes. We also note that McKelson, though he does not give a phonological description, recognises two apicals in his vocabulary with minimal pairs such as ngali 'we dual' and ngarli 'wet ground'. Capell [ 1962 ] , also, describes the language as having two apical series. 2.1.2 Vowels Garadjari has three vowel phonemes, a high front [ i ] , high back [ u ] and a low phoneme [ a ]. There is no length distinction in the vowels and vowels are always short. 2.2 Allophonic Variation Because there is no distinction between a voiced and voiceless series of stops there is variation between these two possibilities. The voiced alternative is taken as basic because this is the most common alternative, and in fact, for the apicals and laminal, is the only alternative. Voiced - voiceless variation is not entirely free because, as stated, in the non-peripheral phonemes we do not get variation but only get the voiced alternative, and with the bilabial and velar stops it is most common to get de-voicing word initially or intervocalically but very rare after a voiced consonant. The exception to this a word final bilabial stop after an [ r ]. ie. [ wurp ] 'plunge' [ gurp ] 'pile' Devoicing in the velar stop is a lot more common than in the bilabial, and in word initial postition occurs about fifty percent of the time. We also find variation between the apical series in certain words, eg. birdi bidi 'hole' though whether this represents actual variation or simply a difficulty in Laves' interpretation of these sounds is unknown. I tend to believe that there are two phonemic apical series and that Laves had some difficulty in distinguishing them. Where we have a homorganic velar nasal plus stop cluster this cluster varies freely with simply a velar nasal, ie. the stop may be deleted. This phenomenon does not occur with the other homorganic clusters and nor does it work the other way, ie. the phonemic sequence [ ng-g] may be reduced to [ ng] , but a phonemic [ ng] may not be followed by the stop to make a cluster. Because Laves used the symbol n for the velar nasal we cannot attribute this to simple inconsistency in the interpretation of n. 2.3 Phonotactics 2.3.1 Word Initial Every word, and in fact every morpheme, in Garadjari begins with a consonant, and we never get a consonant cluter word initially. Only one of the apical series, the alveolar series, may occur word initially, however, we may get either of the apicals suffix initially. It therefore seems we get neutralisation of the apicals in a word initial position. The apicals that do occur word initailly are rare in this position and are greatly outnumbered by consonants from the other places of articulation. Apart from this the only other restriction we have is that the laminal lateral [ lj] does not occur word initially. 2.3.2 Word Final Words most commonly end in a vowel though it is also common to find words ending in a sonorant. We may get any nasal, lateral or rhotic in the word final position with the restriction that the post-alveolar apicals may appear finally only on a pre-auxiliary element of a compound verb ( see section 4.1 ). eg. warl ma- 'to take underneath ?' wurn ma- 'to crack' We also get a few words that end in a stop. With one exception these are words that take no inflectional endings ( note that all inflectional endings begin with a consonant and that consonant clusters with the first consonant as a stop do not occur.), ie. we generally get pre-auxiliary elements of compound verbs ( though they generally occur with no auxiliary ) and some adverbs. eg. djagad djagad = he sneaked along ( ie. sneak sneak ) ginjangka gurag = he jumped up ( in pursuit ) wurb = he plunged gardaga rad = he slept soundly ( ie. sleep hard ) There is one combination of word final consonant cluster that may occur and this is the rhotic followed by a bilabial stop. eg. ngurb = he fell down There are no other final consonant clusters in the data. We do get one noun with a final stop and this is the noun gurb 'pile'. We can classify this as a noun because it occurs with the nominal modifier warayinj 'other' though I have no examples of it taking an inflectional ending and therefore cannot determine what happens here to the consonant cluster that would result. 2.3.3 Consonant Clusters There are fairly strong restrictions on the combinations of consonants that may occur together in a cluster. We get maximum clusters of two consonants, and the first consonant may not be a stop nor a semivowel. As with word initial and word final restrictions we do not get a post-alveolar apical as the second consonant of a cluster though we may get a post-alveolar nasal followed by a stop. We find that the peripheral consonants are more able to form the final consonant of a cluster than the apicals and the laminals but the apicals and laminals are more common as the initial consonant. We can make a chart of the possible consonant clusters as follows : Chart 2.2 Consonant Clusters C2 b d dj g m ng w C1 r + + + + + + + l + + + + + lj (+)? m + (+) n + + + + + + rn + + + ng + nb. ( ) means that the cluster is dubious. This information can be summarised by making the following generalisations. In a cluster C1 C2 we have : ( nb. # means 'does not equal' ) C1 # laminal ( note that the combination [ ljg ] above is dubious ) # stop, semivowel C2 # post-alveolar apical # apical, laminal nasal, rhotic, lateral # y C1+ C2 = nasal + homorganic stop = rhotic + - = lateral + peripheral = post-alveolar nasal + stop # n + w 2.4 Morphophonemic Variation Garadjari has very few rules of morphophonemic change, but these are : (1) nasal --> n / __ dj 1 (2) nj --> ng / __ g (3) i --> a / __ -ya , -ra Rule (1) applies to the ergative and instrumental suffixes, as these both begin with a [dj] when following a consonant. eg. mirling-dju --> mirlin-dju This can be seen to be related to the admissible consonant clusters. If we look at chart 2.2 we see that the only permissible consonant clusters of nasal plus [dj] involve an apical nasal. We must therefore change any sequence of nasal plus [dj] that appears in a derivation to an apical nasal plus [dj] and the alveolar is the more natural or less marked apical and so the natural choice. It is unknown what may happen to the post-alveolar nasal before the ergative suffix ( ie. would this fall in with the other nasals and shift to an alveolar ) because, as has been mentioned, we only get a final post-alveolar sonorant on the pre-auxiliary element of a compound verb and not on nouns, and thus the situation never arises. Rule (2) is also a rule imposed by the permissible consonant clusters, ie. the statement that we do not get a laminal consonant pre-consonantally. Though I have no examples in the data, I would imagine that this rule would apply before any of the consonants that may occur suffix initially. The velar stop is the most common suffix initial phoneme and thus we have examples of this where we do not have it with the other consonants. An example of its appliation is : yalinj --> yaling-gu Rule (3) is a rule of regressive vowel harmony. It applies to the continuous aspect suffix -ri before a suffix with an [a] , ie. the third person plural subject marker -ya and the completive aspect -ra . eg. ya-njari-ya --> ya-njara-ya 1 Note, however, that what we might actually expect here would be the homorganic cluster [ njdj] . The fact that we do not get this cluster may simply be because Laves had difficulty distinguishing a palatal nasal from an alveolar nasal before the palatal. -------------------------------------------------------------------------- end of file