RELATIVACIÓN DE LA INDEPENDENCIA
4. LA DECISIÓN FINAL
In Nakkara there are significant generalisations to be made about the length of stops at morpheme boundaries. Firstly, a large percentage of long stops occur at word-internal morpheme boundaries. Considering the polysynthetic nature of Nakkara words this is an important point. In fact, a quick count of long stops (excluding those in syllable-final position) in a conversational text has shown that in the first two hundred sentences, a staggering 70% of long stops occurred at morpheme boundaries; only the remaining 30% were morpheme-medial long stops. In other words, given an understanding of morpheme boundaries in Nakkara, there is a significant degree of predictability of stop length. Indeed, it is possible to make three far-reaching generalisations about stop length at morpheme boundaries.
1. Root-initial stops in word-medial post-vocalic position will always be long.
2. Stop-initial suffixes always have an initial long stop.
3. Reduplicated and compound word-forms represent special cases with regard to root morpheme boundaries. (See 2.4.1.3 to 2.4.1.4 below). Nevertheless they normally fit into either of the categories described in 1. and 2. above, and stop length at their internal boundaries is predictable to this extent.
The lengthening of root-initial stops in word-medial position (as described in the first generalisation above) is also referred to as stop alternation. Comparison with the major type of reduplicated word-form (Reduplication Type A) suggests that these root-initial stops are
Sections below examine the phenomenon of stop alternation in Nakkara as well as the occurrence o f stops at various types of word-internal morpheme boundaries, including those in reduplicated and compound word-forms. The coincident factors of word stress and accompanying vowel length are also mentioned, with the possibility that these factors might play a greater role in conditioned stop length if the morpheme boundary trigger were not as regular and consistent. All of these different m orphological types are discussed in an attempt to capture and account for the degree o f predictability o f stop length at morpheme boundaries.
Section 2.4.1.1. exam ines the m ost com m only found stop alternation in verbs and nominals. Stops occurring at the boundary of stem and suffix are discussed in 2.4.1.2.
2.4.1. Stop Alternations in Nakkara
Stop alternation occurs most commonly across the morpheme boundary o f prefix and verbal or nominal root. Specifically, a word-initial stop, which is always short, occurs as a long stop when forced by a prefix to occupy word-medial position. Alternation of this type or similar has been docum ented for other languages in the Arnhem Land area (cf. Waters 1983 for Djinang, M cKay 1984 for Ndjebbana) but it is not nearly as productive a process in either o f these other languages.
2.4.1.1. Stop Alternations in Verbs and Nominals
Lengthening o f a word-medial morpheme-initial stop is a common process in Nakkara and is quite productive, with only a few exceptions. Some examples follow:
(1) i) a) Qa' b : ud a Qa 7 hit him!her' Qa +b:uda - Qa
1M(A) -4-hit -cyr 3M (0)
but b) ’ b u d a Q a 's/he hit him' 0 +buda - Qa
3M(A) -4-hit -QT 3 M (0 )
ü)
but
üi)
but
a) nurba'giaramada 'you lot got up' nurba +g:arama -$a
2AUG(S) +get up -PC/T
b) 'gaiamada 'he got up' 0 +garama -$a
3M(S) + getup -PC/T
a) na'greyada 'country, clan estate'
na +geyada
NOM +camp
b) 'geyada 'camp, "home-camp"'
An obvious question arising from stop alternation betw een tw o otherw ise identical morphemes is: can we establish an underlying form? Given the previously mentioned environments for neutralisation o f stop length we can assume that this accounts for the short stops in word-initial position in ib), iib) and iiib). This m ight suggest that the underlying form is that found in word-medial position. Along these lines of argument the underlying form o f the verb root 'to hit' would be /bruda/ and 'to get up' /g:aiama/, while the underlying nominal stem in iii) would be /greyada/, that is, all would have an initial long stop. Nevertheless, the existence of a similar phenomenon in the major type of reduplicated word-form militates against this proposition. Consider the examples o f reduplication type A listed under (2) below:
i) 'garagarawab:a gara-gara+wab:a 'all along’ ü) na'g:aragara na+gara-gara 'long way' iii) n'badabada n+bada-ba<Ja 'Barramundi' iv) na'g:emgera na+gera-gera 'Sand Crab' v) na'g:obagoba na+goba-goba 'Ghost Crab' vi) na'g:uyaguya na+guya-guya 'th ie f
Focusing particularly on examples (2) i) and ii) we see the same type o f alternation as in (1) above. If, by analogy, the underlying form of the reduplicated morpheme segment has an initial long stop, we would expect the underlying form to be reduplicated. Instead, the second occurrence o f the segment consistently shows a short stop. We can therefore assume that in alternations such as these the underlying stop phoneme is short. There are coincident factors involved in stop alternation in Nakkara, (e.g. stress patterning), but the regularity and productiveness of the process in this language means that it can be accounted for in terms o f the morpheme boundary alone, as in the following rule:
ROOT
[i.e. a consonantal stop segment is created post-vocalically, identical to the following root-initial stop.]
Note in this regard that prim ary stress co-occurs with the initial syllable of the root morpheme. In fact, in my earlier analyses o f Nakkara phonology, primary stress was put forward as the conditioning factor for stop alternation. In neighbouring Ndjebbana where the alternation is not as productive, primary stress placem ent is used as an appropriate conditioning factor for stop length alternation, (cf. McKay (1984).)