RELATIVACIÓN DE LA INDEPENDENCIA
3. EN BUSCA DEL ARCA PÉRDIDA
3.4. ENDURECER LA AUTOREGULACIÓN
An elementary step in assessing the phonological status o f short and long stops in Nakkara has involved looking at their distribution. Certain trends are apparent from this approach, the m ost significant o f which is that long and short stops are not in totally contrastive distribution. The Table below illustrates this clearly:
# _ v _ v N__V L__V R__V y _ v _ # Bilabial b b / b : b b / b : b / b : b / b : - Apico- alveolar d d : d d - - d : Post-apico
a / a :
-a:
a:
a:
alveolar Lamino- jj / r -
J
J / J :
? •J:
T- palatal Dorso- g g / g : g g / g : g: g : g: velar[# = word boundary; V = vowel; N = nasal; L = lateral; R = rhotic; y = semivowel /y/;]
Within the word, stop length is only contrastive in medial positions. The most straightforward distributional restrictions are found at word boundaries and also at syllable boundaries and these are discussed below. We then look more closely at stop distribution within word boundaries.
2.3.3.1. Distribution of Stops at Word Boundaries
In word-initial position production of stops can be either voiced or voiceless but the stop is always short (indeed, sometimes fricated). In word-final position there is someWsno distinctive release. Nevertheless, no voicing is involved in production and a word-final stop is always considerably longer than one in word-initial position. In summary, only short stops occur in word-initial position; in word-final position, only long stops occur.
2.3.3.2. Distribution of Stops at Syllable Boundaries
Syllable-initial stops are always short and syllable-final stops are always long (but generally somewhat shorter than word-final stops). In terms of voicing characteristics, syllable-initial stops are sometimes voiced (but voicing may fluctuate) and syllable-final stops are usually devoiced. Therefore, a parallel between distribution of stop length at word and at syllable boundaries exists in Nakkara.
Inclusion of a distributional parallel of this type has not been a feature of analyses of other languages in the area. There has never been any argument about the type of stops that occur syllable-finally in these languages. They are always perceived as being very similar in phonetic character to word-final stops. There have been, however, differences in interpretation for syllable-initial stops, depending on the position of the syllable in the word and also on the overall interpretation adopted for resolution of the stop contrast problem. In the section following we look closely at data showing heterorganic stop clusters. These are the most suitable examples for determining stop length at syllable boundaries.
2.3.3.3. Heterorganic Stop Clusters
The value of using examples of heterorganic stop clusters for establishing syllable boundary stop distribution is that these clusters consist of two definite segments, the first of which is, quite clearly, phonetically long.
Glasgow and Glasgow (1967) and McKay (1980) have argued that heterorganic stop clusters are always devoiced, which suggests that each stop in the cluster becomes voiceless. McKay (1980:344) makes a link between voicelessness and fortisness and has extended the former statement to claim that heterorganic stop clusters are, in addition, fortis. Nevertheless, we have established that voicing alone is not a reliable phonetic parameter and that without precise phonetic specification the terms fortis/lenis are not particularly useful labels.
If an analysis of heterorganic stop clusters and/or any other occurrence of stops relies on voicing as a predominant distinctive feature of the contrast, syllable-final stops can always be described as voiceless, and it would be easy to assume, as McKay 1984:107 does, that a stop immediately following this segment would also be voiceless due to progressive assimilation. If, however, we agree that the principle feature distinguishing stops is duration, a process of assimilation cannot be assumed to be operating between the two stops. That is, it does not necessarily follow that the length of the second stop in a cluster will be predetermined in the same way. I argue here that for Nakkara there is a significant parallel between the duration of stops at word and at syllable boundaries, and I therefore focus on the phonetic nature of stops in the language.
My own Mingograph readings consistently show significantly long duration for heterorganic stop clusters in Nakkara. (See 2.3.1.1.) No consistent readings for voicing were apparent, but this correlates with word-initial stops and the apparent variation implies, I suggest, voicing fluctuation. This phenomenon has turned up elsewhere. For example, McKay (personal communication) has said that analysis of the spectrographic readings of heterorganic stop clusters in Rembarmga showed no consistency with voicing.
That heterorganic stop clusters have consistently long duration is to be expected if we think of the cluster quantitatively as the sum of two individual segments. Nevertheless, it is interesting that the duration of the combined stops in these clusters is equivalent to, and certainly no greater than, long stops in similar environments.
This is a significant point: there is a quantitative equivalence between two non-identical stops flanking a syllable boundary and an intervocalic long stop. Indeed, although the
heterorganic cluster consists o f two definite segments, the first o f which is phonetically long, the duration of the sum o f the two stops is no longer than that of an homorcjan < c long stop, which we might reasonably expect to be a single segment.
I have observed that for heterorganic stop clusters the syllable-final stop has no independent release burst; rather, the release is heard at some point during production of the stop follow ing (syllable-initial). These perceptions w ould seem to match Catford's (1977:221) description o f heterorganic 'close transition' wherein he claims,
'there is no articulatory continuity, but there is articulatory overlap... the articulatory stricture for the second consonant is formed before the stricture for the first is released... the second articulation is formed well before the first is released.' The second stop is released almost immediately after formation o f the obstruction and is then coloured by voice onset o f the vowel immediately following. It is this, I suggest, that sometimes gives the impression o f voicing in syllable initial position. Given that duration of heterorganic stop clusters is equivalent to duration of long stops occurring in intervocalic position and also that release of the syllable-initial stop appears to be almost instantaneous, there is no reason to posit (as does McKay (1980), (1984)) exceptions to the distribution of stops at syllable boundaries. Cf. McKay (1980:343) and (1984:107) respectively:
'In general syllable initial stops are lenis (except in intersvllabic stop clusters! and syllable final stops are fortis.' (my emphasis)
'When a syllable final and a syllable initial stop come together within a word creating a cluster of two stops (whether hetero-organic or homorganic) the syllable initial stop assimilates to the syllable final stop to become voiceless and fortis.’
I argue that the syllable-initial stop in the cluster is short, and that voicing of this stop, although difficult to perceive clearly, is variable, implying non-significant fluctuation. On these grounds I posit a distributional parallel o f stop length at word and syllable boundaries. Hence, word-initial and syllable-initial stops are short and sometimes voiced (but voicing may fluctuate). With word-final and syllable-final stops the release is delayed somewhat and duration is considerably longer, long stops are usually voiceless.
There is an additional parallel between heterorganic stop clusters and long stops that is significant: there is a constraint on occurrence of both stop clusters and long stops word- initially.
2.3.3.4. Distribution of Stops in Word-Medial Position
In the two types of environment discussed above (word boundaries and a restricted set of syllable boundaries), there is clearly complementary distribution of short and long stops, suggesting allophonic status for one o f the series. Nevertheless, when we look at the distribution o f stops in word-medial position (cf. Table 2.3.3.) we see a range o f contrasts, which without further investigation confound any attempt to eliminate one o f the series of stops from the phoneme inventory.
Distributional analysis o f stops in word-medial position in Nakkara is best tackled by first establishing a clear distinction between morpheme-medial occurrence and occurrence at morpheme boundaries. In 2.3.3.5. below the discussion is c o n fin e d to phonology proper, drawing on examples showing morpheme-medial occurrence o f stops. It is for long stops in this environment that a geminate interpretation is proposed. Morphophonemic alternation o f long and short stops and other predictable occurrences of long stops are discussed along with the range of morphophonemic processes evident in Nakkara, in 2.6.2.ff.
2.3.3.5. Morpheme-Medial Long Stops
Morpheme-medial occurrence of both long and short stops represents the crux of the stop contrast problem in Nakkara. As the examples below suggest, there is an apparent situation of parallel distribution word-medially.
- c - - C : -
'boiaba 'to paint / rub', 'bornbia 'to be in pain' 'buda 'to hit, kill’, 'bud:a 'y ab b y ' ’gagada 'on top of, high' '$ag:ag:a 'b a b y '
An advocate o f the segmental solution could at this point justify expansion of the phonemic inventory (thereby sacrificing phoneme economy) by pointing to these and other contrastive pairs. In doing so, distributional restrictions and the inventory o f syllable types would also have to accommodate long stops in syllable-initial position. Apart from these compromises, the phenomenon o f contrasting stops would then cease to be a problem.
For a prosodic analysis the proponent would have to call on some suprasegmental feature that accounts for the contrast. W ood (1978) and more recently W alker (m.s.1984) have presented some very good evidence for this interpretation for the Yolngu languages of north-east Arnhem Land, although their solutions do not help solve this phenomenon in all Yolngu languages.2 This prosodic interpretation relies on the existence of a glottal stop in the language which in certain circumstances creates a "fortis syllable"; this then triggers stop length in the following consonant segment. Because of the absence o f the glottal stop in N akkara (and also N djebbana (Kunibidji) and Burarra), or o f any other obvious conditioning factor in this environment, the prosodic interpretation as it stands cannot be motivated for these non-Yolngu languages. Nonetheless, in sections 2.6.1. to 2.6.2. below we mention another prosodic feature, stress/vowel length, as a coincident factor involved at the morpho-phonemic level in relation to predictable stop length at morpheme boundaries.