Volume 75 1966 > Volume 75, No. 4 > Narak: language or dialect?, by Edwin A. Cook, p 437 - 444
NARAK: LANGUAGE OR DIALECT? 1
Narak speakers are one of several language groups in the Jimi River Valley of the Jimi River Sub-District of the Western Highlands District, Territory of New Guinea. They number about 3,000 and are distributed primarily on the southern flanks of the Bismarck Mountains as well as on both banks of the middle Jimi River between lat. 5° 20-40″ S. and long. 144° 30-50″ E. The Administration Patrol Post at Tabibuga is in their territory. They are shifting cultivators utilizing an altitudinal range from 1,800′ to 6,000′. The terrain is exceedingly rugged, there are no flat valley floors. The primary cultigen is the sweet potato which is supplemented by pandanus, taro, yams, bananas, sugar cane and other crops. They are avid pig raisers and hunting and gathering are relatively unimportant. At the most they have known steel implements for 15 to 25 years and less in some of the more inaccessible areas. Under the Australian Trusteeship Administration there have been no large scale wars for several years, the last occurring in 1956.
Normally, ethnographers simply state that the language of the people they have studied is language ‘X’. Although I have been tempted to follow this example on occasion, I find it scarcely possible due to the differences of opinion expressed in the works of those few authors who have published on the languages of this area.
One of the difficulties in administering the peoples of the Jimi River is their tremendous linguistic diversity. In the early 1950's, Patrol Officers often found it necessary to employ two interpreters in order to communicate at all with the newly contacted natives. The Patrol Officer would translate from English into Neo-Melanesian for the first interpreter, who would then translate Neo-Melanesian into his own language, say language ‘A’, for example. The second interpreter, understanding language ‘A’, would then translate into language ‘B’. Though language- 438 - 439
‘B’ would not be the proper local language or dialect of the newly contacted natives, they would, nevertheless, usually be able to understand ‘B’ and to communicate with the second interpreter by speaking ‘C’, their own language or dialect. This process had led to many misunderstandings between the Administration and the non-Neo-Melanesian speaking population.
As a result of this generally confused situation, the Administration requested the Summer Institute of Linguistics, New Guinea Branch, to undertake a linguistic survey to include the Mount Hagen Sub-District of the Western Highlands District. This survey was conducted by Gordon Bunn and Graham Scott between August 1 and September 12, 1961. 4 The data are sketchy, since in most instances comparison of languages was based on a sample of 50 lexical items. 5 They spent no more than two weeks in the Jimi Valley and, considering the task at hand, it is remarkable that they accomplished as much as they did.
Bunn and Scott distinguished a maximum of five languages within the boundaries of the Jimi River Valley. These five are: Medlpa, Karam, Maring, Narak and Gandja. 6 In comparing this listing with Wurm's and Wurm and Laycock's listings, 7 several differences in terminology appear (Table I).
NAMED LANGUAGES IN THE JIMI RIVER VALLEY
Both Wurm and the Bunn and Scott team computed percentages of cognates between most of these languages. Table II compares their computations. Since it is known that Bunn and Scott used a greatly shortened list, and that Wurm used a much longer list, I feel that Wurm's figures are much more reliable. However, at the same time, I suspect that longer lists yet would produce even higher percentages of cognates.- 440
PERCENTAGES OF COGNATES 8
One of the more interesting points to be noted in this Table is Bunn and Scott's failure to distinguish Narak from Kandawo; i.e. their Gandja. Since they differentiate one language from another on the basis of sharing less than 85% of their basic vocabularies, 13 and since, according to Wurm, Narak and Kandawo share only 61% of their basic vocabularies, then we may assume that either Bunn and Scott's truncated list was not sensitive to this difference, or that they did not touch on places in the Jimi where this difference would have been revealed. They do concede that there are two dialects of “Gandja” and that, “. . . mutual understanding is very limited due to the fact that they almost border on being two separate language groups.” 14 It is difficult to determine whether they intend “separate” to mean linguistic divergence or spatial separation. Bunn and Scott's work was also hampered by the fact that at the time of the survey there were virtually no Neo-Melanesian speakers in the Jimi River area. 15
The problem here is whether the speech of a group of natives should be termed a dialect or a language. As Wurm and Laycock note, this is a problem of deciding when the speech of two communities is mutually intelligible or non-intelligible. Wurm had originally utilized Swadesh's criterion that the speech forms of two communities are to be considered dialects when they share more than 80% of their basic vocabularies. On this basis, Wurm established a typology of Highland languages. The East New Guinea Highlands Stock (741,000 speakers), whose members share 12-28% of their basic vocabularies, is composed of five language families, sharing 28-81% of their basic vocabularies. One of these five - 441 is the Hagen-Wahgi-Jimi-Chimbu Family (282,000 speakers). The Jimi Sub-Family is one of the four Sub-Families of this Family and is composed of three languages: Yoadabe Watoare (Maring)—3,200 speakers, Narak—6,150 speakers, and Kandawo—6,750 speakers. 16 The percentage of cognates in basic vocabulary between Maring, Narak, and Kandawo is as follows 17:
However, Wurm and Laycock both felt that, “. . . the percentage figure quoted (of 81%) corresponds to the reality of the borderline between mutual intelligibility and unintelligibility only in rare instances.” 18
These two authors cite four methods, derived from Voegelin and Harris, 19 for determining the extent of mutual intelligibility: (1) the opinion of the informant, (2) counting cognate lexemes, (3) comparison of phonology and morphology, and (4) testing the informant with a text to translate, 20 Wurm and Laycock found that when speakers shared between 60% and 70% of their basic vocabularies there was some mutual intelligibility. But when the percentages began to drop below 60%, intelligibility was then more dependent on congruence in phonological and morphological structures. 21 Relying on the opinion of the informant as a criterion of relatedness is, they say, the weakest of the methods. 22 Testing the informants by having them translate texts is the most reliable, but is also the most time consuming and is seldom attempted by ethnographers in the normal pursuit of their work.
There is also the additional consideration of the phenomena of language chains. 23 That is, of three languages A, B, C, distributed lineally in space, where ‘+’ means dialect and ‘—’ means language boundary, you may have the arrangements: A-B-C, A+B-C, A-B+C, and A+B+C. In the event of intrusion you might also have A+C-B. Taking the 81% figure, Maring, Narak, and Kandawo, which are distributed more or less lineally in space, are of the form A-B-C. 24 Using the 60% minimum figure, Maring, Narak, and Kandawo are of the form A-B+C, 25 using the 50% figure reduces all three to dialect status, i.e. A+B+C. 26 Since Wurm and Laycock were not peculiarly interested in the Jimi Sub-Family, the problem was left unresolved.- 442
However, in a recent article, Wurm uncritically presents all three solutions. 27 First, in listing members of the Central Stock by languages within their respective Families, Wurm lists Maring, Narak, and Kandawo as three separate languages. 28 Secondly, citing his own and Laycock's previous publication, 29 Wurm regroups these languages so that Maring retains the status of language but Narak and Kandawo are classified as dialects of a second language, Narak-Kandawo. Although Wurm does not explicitly state his criteria for this rearrangement, he must have used the 60% cognate figure, since a lower figure (e.g. 50%) would have resulted in the inclusion of Maring as a third dialect. Third, if we consider similarity and dissimilarity in structural features (phonological and morphological), then the data in the tables presented by Wurm 30 supports a Maring, Narak, Kandawo relationship of the form A+B+C, i.e. all dialects, since each agrees with the other two in either the presence or absence of all twenty-one of the listed features. I would suspect that this probably reflects a general dissatisfaction with arbitrary criteria for establishing the distinction between language and dialect.
My own data, obtained by asking the informant, supports the arrangement A-B+C. That is, informants are quite certain that Maring is an altogether different language. Informants from both Togban and Kwiop (Map numbers 43 and 42 respectively) each maintain that the other speaks a different language. On the other hand, as one progresses upriver toward Bubgile (Map, number 37) or to the south toward Korendiu (Map, number 18), informants maintain that the next speech community is either identical with their own or only slightly different, a difference which does not seriously restrict the transfer of information. Many natives of these small speech communities are extremely sensitive to even the most minor of dialect differences and, within a limited spatial range, a perceptive informant can tell you almost precisely in which village an individual was raised as a child. Dialect variations constitute one source of humour, and the speech of natives from slightly or greatly varying speech communities is often the subject of exaggerated mimicry by local inhabitants. My own limited command of Narak was achieved at Kwiop. Later, while residing at Warames (Map, number 17), I conducted an informal experiment. With myself as narrator, I taped several short texts and played them back to informants who were unaware that they were listening to my voice. They were invariably able to place the speaker's speech community as Kwiop.
The accompanying map shows the language boundaries within the Jimi River valley which were achieved by asking informants. The boundary through Korendiu is further supported by Marie Reay. She notes that the language of the Kuma, located near Minj in the Middle Wahgi, is known as “Yoowi”. This language, she states, is spoken as far north as “Korendyu” in the “Jimmi” where it gives way to - 443 “Yoonerag”. 31 Further, the Reverend Father Joseph McDermott, S.V.D. (stationed at Ambulla, Number 29 on the map) speaks Chimbu (Bunn and Scott's “Kuman”) 32 and states that he has no trouble either in communicating or understanding the native language around Monggum (Number 28). Brookfield and Brown note that Chimbu speakers live on the north side of the Bismarcks. 33
This additional information has not, of course, resolved the problem entirely, but only an extended linguistic analysis could produce better results. There are many potential faults with the ‘ask the informant’ method. Voegelin and Harris point out that, “. . . the informant's perception may or may not agree with the amount of linguistic difference between the dialects in question.” 34 However, whatever features the native speakers may be recognizing when he makes a distinction between two speech communities, he is ultimately referring to the transfer of information, i.e. mutual intelligibility.
It appears to me that of Voegelin and Harris' four methods; the second and third are, in Pike's terms, primarily etic, 35 whereas the first and fourth are emically oriented. Etic analyses are, in general, oriented toward comparative purposes and therefore tend to be utilized primarily as classificatory devices. In contrast, emic analyses represent an attempt to discover categories meaningful to the native speaker, whether these catgories are in the realm of kinship, 36 plants, 37 diseases, 38 or, as I propose, dialect differences. Further, the emic approach is essential when considering the problem of language chains. In studying the relation of one speech community to another, as seen from the viewpoint of the native speakers, we may be able to intelligently perceive the organization of dialects into languages and thus bring to a close endless quibbling over 10 or 20 percentage points.- 444
1 The research upon which this article is based was carried out under the tenure of Predoctoral Fellowship MF-11,543, and Research Grants M-4895 and MI-04895-02s1, from the National Institute of Mental Health. United States Public Health Service. This manuscript has benefited from the helpful criticism of Bruce Biggs, Ralph Bulmer, Andrew Pawley, A. P. Vayda, George Grace and Leopold Pospisil, though they are not necessarily to be associated with the viewpoint I have maintained.
2 Language groups are as distinguished by informants whose places of birth and continuing residence are near Census Points 6, 10, 17, 18, 22, 26, 29, 37, 40, 42, 43, 44, 46, 52, and 61. [Boundaries between Western groups as here defined do not coincide with reported boundaries of the languages known as ‘Enga’ and ‘Medlpa’ in the general literature; as far as the Editor is aware, there are no Enga speakers within a day's walk of the Moglpin River].
3 69 Government Rest Houses or Census Points are noted on the map by the small numbers. A listing of these with their altitudes and populations may be obtained by writing to the author.
4 Bunn and Scott 1962. Permission to quote from Bunn and Scott's work was kindly granted by Karl Franklin, Director, New Guinea Branch, Summer Institute of Linguistics, Ukarumpa, Eastern Highlands District, Territory of New Guinea.
6 Ibid.: map following p. 8.
7 Wurm 1961, 1962, 1964: Wurm and Laycock 1961.
8 The language labels used in the table follow Wurm 1961.
9 Since I am using Wurm's distinctions, Narak and Kandawo are separate columns. Bunn and Scott do not segregate Narak from Kandawo: therefore the percentage of cognates with any other language will be identical in both columns.
10 Bunn and Scott's percentages, 1962.
11 3. Wurm's percentages, 1961.
12 No figure given by Wurm (1961:21).
13 Bunn and Scott 1962:3.
14 Ibid.: 5.
16 Wurm 1961:18.
18 Wurm and Laycock 1961:129.
19 Voegelin and Harris 1951.
21 Wurm and Laycock 1961:134.
27 Wurm 1964: Passim.
29 Wurm and Laycock 1961.
30 Wurm 1964:85.
31 Marie Reay (1959:1) speaks of the language near Minj as Yoowi and states that, “. . . the language changes from Yoowi to Yoonerag at Korindyu, about midway between the heights of the Wahgi-Sepik Divide and the Jimmi River.” Local informants maintained, however, that the language, “Yoowi”, was split in the Jimi into “Yu” and “Wi”. From distributional evidence it would appear that speakers of Wi are later migrants to the Jimi than speakers of Yu though this is only speculative. It seems to me that “Yoonerag” may be composed of the morphemes yua (man) and narak (what). A loose translation (from the point of view of the Minj people) into English would render this as, “The ‘what?’ man”, perhaps implying that in response to Minj speech, Jimi people past Korendiu would invariably reply with narak?, i.e., “what?”.
32 Bunn and Scott 1962:5.
33 Brookfield and Brown 1963:3.
34 Voegelin and Harris 1951:324.
35 Pike 1954:8-31.
36 Lounsbury 1964.
37 Conklin 1962.
38 Frake 1961.