Volume 70 1961 > Volume 70, No. 1 > Migrations implied by relationships of New Britain dialects to central Pacific languages, by Ward H. Goodenough, p112-126
                                                                                             Previous | Next   

- 112
MIGRATIONS IMPLIED BY RELATIONSHIPS OF NEW BRITAIN DIALECTS TO CENTRAL PACIFIC LANGUAGES

Dr. Goodenough, of the University of Pennsylvania, is well known for his studies of social organisation, but in this article he makes a contribution to Oceanic linguistics. Dr. Goodenough directed a programme of field research in New Britain from 1954 to 1956.

1. My concern in this paper is with the Nakanai dialects of the north coast of central New Britain in Melanesia and their relation to Central Pacific languages.

1.1. George Grace has recently pointed out that of the languages examined by Dempwolff, “Fijian, Sa'a and the Polynesian languages show an impressive number of innovations in common.” 1 He has designated this obvious subgroup of the Austronesian or Malayo-Polynesian languages as “Eastern Austronesian”. In this group he includes a number of other Pacific languages, including Rotuman, spoken on a small island near Fiji. Within Eastern Austronesian, Grace argues that Fijian, Rotuman, and Polynesian constitute a further subdivision, more closely related to each other than to any other languages. This part of his thesis may require some modification, as I shall attempt to show, 2 but Grace's work documents the very close relationship between these three languages. They are constituent of what I am here calling Central Oceanic in order to distinguish it as a subdivision of the more extensive Eastern Austronesian. Central Oceanic may be tentatively equated with what Grace has elsewhere 3 called the “New Hebrides-Banks” group within “Eastern Malayo-Polynesian” (Eastern Austronesian) and, more specifically, his subgroup “b” within this, which includes some New Hebrides languages as well as Rotuman, Fijian, and Polynesian.

1.2. Nakanai speakers occupy a one hundred mile stretch of coast along the north shore of central New Britain. 4 At one point, Nakanai speech extends inland, but it is otherwise coastally distributed. This is in - 113 marked contrast with most other languages in central New Britain, which extend deeply inland, some all the way to the south coast. From its distribution I infer that Nakanai has spread at the expense of older languages, still represented inland, its speakers becoming linguistically dominant over a wide area. Comparative evidence suggests that this spread did not come from within New Britain itself, but was due to the arrival on New Britain's shores of linguistically alien people.

2. If Nakanai can be reasonably presumed to be a relative newcomer to New Britain, we then have to consider from what direction it might have come. There are, of course, any number of possibilities, but there are two major postulates about Pacific Island migrations which commend themselves for immediate consideration.

2.1. West to East population movement from Southeast Asia. There is good reason to assume that in the course of time successive language groups have worked their way eastward from Southeast Asia through Melanesia. Some of these groups, presumably those who started east-ward earlier, ultimately reached the Central Pacific. This being the case, those languages in Melanesia whose distribution suggests that they belong to an older linguistic stratum should be more closely related to the Central Pacific languages than those which appear to belong to a younger stratum, while the latter should show closer relationship with more westerly situated language groups.

2.2. Secondary east to west population movements from the Central Pacific. From his review of the evidence, Andrew Sharp has concluded that castaways and involuntary migrants were common before European contact, that few of them had an idea where they were going, that trusting to favourable winds and currents they tended to move from east to west in the great majority of cases. 5 As soon as the primary west-to-east drift led to settlement of islands farther out in the Pacific, these islands became centres from which castaway groups were fed back westward upon the already populated regions or to hitherto unpeopled islands which were now in the line of drift from the newer settlements. Evidence that this reverse direction of population movement has been going on in fairly recent times is conclusively demonstrated by known instances of castaways, by remembered incidents of only a few generations ago, and by the Polynesian outliers in Melanesia and pockets of Polynesian speech in the larger Melanesian islands. It seems reasonable to assume that this process has been an active one since the Central Pacific was first populated, and that the Polynesian outliers represent only its later phases. Before Polynesian became differentiated out as such from other early central Pacific dialects, speakers of these old dialects must have been peppering the eastern shores of Melanesia with landings, in some cases managing to maintain their separate linguistic identity, and occasionally giving rise in time - 114 to a proliferating language group. If this is the case we should expect at least some of the newer linguistic strata in Melanesia to show a closer relationship to the Central Pacific languages than is shown by the older strata. If we can find languages which show this pattern of relationship, they will provide confirmation of this hypothesis that the Central Pacific has played an important secondary role in the peopling of Melanesia.

With this in mind, we return now to Nakanai. The first problem is its relation to other languages on New Britain and to Grace's Eastern Austronesian group.

3. The languages on New Britain most closely related to Nakanai appear to be those spoken on the Willaumez Peninsula and, possibly, the group of dialects known as Extended Mengen, on which I have no data. The other languages for which I and my colleagues have word lists, and these include all those spoken inland behind the Nakanai villages, appear prima facie to be more distantly related. Even the Austronesian status of one (Uasi) is highly doubtful.

FIGURE 1
Distribution of Languages and Dialects in Central New Britain.

3.1. For example, the two languages known as Mangseng and Kapore (Banaule) are far from being mutually intelligible. They show considerable difference in phonology, morphology, and lexicon; yet it is evident that they share features which make them more closely related to each other than either is to Nakanai. A crude application of lexico-statistical methods 6 (the quality of the data permits nothing better) - 115 has provided further confirmation. Using the 215 items on TRIPP Word List I, we get the following percentages of shared cognates with similar meanings. 7

Mangseng and Kapore (192 items) 32%
Bileki (Nakanai) and Kapore (190 items) 15%
Bileki and Mangseng (210 items) 12%
Fijian and Mangseng (203 items) 13%
Bileki and Fijian (211 items) 28%
Bulu (Willaumez) and Fijian (206 items) 25%
Bulu and Bileki (211 items) 39%

3.2. These figures provide statistical support for what inspection alone (see Tables 1 and 2) 8 suggests. There is reason to assign Nakanai and Willaumez to Grace's Eastern Austronesian. But this does not provide conclusive evidence as to migrations. Nakanai could, along with some other languages in Melanesia, represent remnants of the linguistic group which got out into the central Pacific area. Eastern Austronesian may, when all the evidence is in, encompass a great many languages in Melanesia, as Grace has suggested in a paper reporting the tentative conclusions of his linguistic survey. 9 Fortunately, Grace's recent study 10 of Central Oceanic, as I am calling it, makes it possible for us to assess the position of Nakanai with greater precision.

4. Nakanai and Central Oceanic. According to Grace, Fijian, Polynesian, and Rotuman constitute a distinct group, Central Oceanic, each being more closely related to the other two than to any other language. This conclusion rests on a number of innovations evidently shared by Polynesian with either Fijian or Rotuman, or both. Grace notes that Nakanai may share one or two of these innovations. Let us look more closely.

- 116
TABLE 1
SAMPLE VOCABULARIES
English Mangseng Kapore Nakanai (Bileki) Willaumez (Bulu) Fiji
‘house’ vel bele luma vale
‘butterfly’ vov bavop bebe bembembe beebee
‘rain’ vit biti hura kandavu uca
‘sugarcane’ to to tovu dovu
‘egg’ otel kotel, koten hatotolu hatelu-ka yaloka
‘stands’ mit ru-mit magiri paiti tuu
‘liver’ ete-po-k (my) kote hate xate yate
‘vomit’ mutmut mututu kalalua lualua lua
‘bird’ men men anahaka malu manu manumanu
‘my ear’ teliNe-k telene-k gavusa-gu taliNa-Ngu daliga-qu
‘louse’ Nut nut utu xutu kutu
‘see’ De-pol i hilo vuti raica, kunea
‘drink’ DiN, Din in liu inu gunu
‘dies’ rin rin, rini peho, mate mate mate
‘his arm’ i-meni pu-mini lima-la lima-na liga-na
‘sleeps’ kulkul konkon mavuta makalu moce
‘fresh water’ Dei neki lalu nalu dranu
‘earth oven’ um kum humu lovo
‘soul’ öni-k (my) pu-kani (his) kalulu, hitu yalo-na (his)
‘leg’ eve-k (my) keve-k (my) vaha-gu virigo yava-na (his)
‘belly’ öpö-k (my) pu-kapa (his) tia Nala kete-na (his)
‘eats’ an ken, kenen ali xaxani kani
‘night’ pe-mlik le-migili logo boNi bogi
‘hear’ noN-pol nono, lono lolo loNo-n-i rogo
‘I’ alö e-i-au au koi au, au
‘one’ omöle, omwale ke i-sasa tara e dua
‘two’ aini lu i-lua lua e rua
‘three’ me, mee miok i-tolu tolu e tolu
‘four’ pinäl, penäl päläl i-vaa vaa e vaa
‘five’ lim limi i-lima lima e lima
‘six’ limae hoten ke i-uolo xono e ono
‘seven’ limai hoten lu i-vitu vitu e vitu
‘eight’ lime, limee hoten miok i-ualu alu e walu
- 117
‘nine’ limbinäl hoten päläl uala-siu rio e ciwa
‘ten’ mule, muule luhul ke sa-vulu ra-Na-vulu sa-ga-vulu
‘left (hand)’ kereNa akos mauli, meru mauli mawii
‘nose’ u-Ne-k (my) noho-limi ma-isu iru ucu
‘cry’ taN, taNaN tali taitali taNi tagi
‘cooks’ mök-i (it) mok i (it) gutu-a (it) gutu-i (it) tavuna, vavia
‘bite’ ano-i kalo i ala, ale xalatata-i
‘bite’ atet-i gorapu gati pandamu kata, kati-a
‘fish’ silaN le ia ixa ika
‘tree’ uön, won kä, ke obu xai kau
‘leaf’ uön-Doi kä-koki gala-lau lalau drau
‘stone’ um, anral kum uati kendo vatu
‘sea water’ Dev mago ndari taci
‘tongue’ muma manapa kalamea tambele yame
‘road’ soveNalo gilikua gauru ndala sala, gau-ni-sala
‘blood’ mla gogoru kasoso ndara draa
‘heart’ teli-k po (my) pu-le kum (his) bua mbua uto-na (his)
‘neck, back of head’ kei-k (my) hi kisu kindu kesu
TABLE 2
AFFIXED POSSESSIVE FORMATIONS IN MANGSENG, KAPORE, BILEKI, MERAMERA, WILLAUMEZ, AND FIJIAN
  Mangseng Kapore Meramera (Nakanai) Bileki (Nakanai) Bola (Willaumez) Fiji
1st sg. meni-k (arm) palimi-k (head) tabu-gu (nose) la-maisu-gu (nose) mata-Ngu (eye) -qu
2nd sg. menu-m palumu-m tabu-mu la-maisu-mu mata-mu -mu
3rd sg. (m) i-meni pu-palimi tabu-na la-maisu-la mata-na -na
3rd sg. (f) i-meni ti-palimi tabu-na la-maisu-la mata-na -na
3rd sg. (n) i-meni ä-palimi tabu-na la-maisu-la mata-na -na
1st du. inc. to-meni oraini su-palimi-ri tabu ne-qetau, ne-itau la-maisu-galua, -gala mata ne-tarua -daru
1st du. exc. ? mese-palimi-mem tabu ne-qamilu la-maisu-milue, -mila mata ne-maria (?) -irau
2nd du. o-menu-mu oraini musu-palumu-mu tabu ne-qamulu la-maisu-mulua, -mula mata ne-marua -mudrau
3rd du. o-meni oraini su-palimi-si tabu ne-sulu la-maisu-girua, -gira mata ne-ria -drau
1st pl. inc. to-meni to-palimi-mem tabu ne-itou la-maisu-gatou mata ne-xita -datou (tr.), -da (pl.)
1st pl. exc. ? me-palimi-mem tabu ne-qamiteu la-maisu-miteu mata ne-mia -itou (tr.) -imami (pl.)
2nd pl. o-menu-mu mo-palumu-mu tabu ne-qamutou la-maisu-mutou mata-ne-mua -mudou (tr.), -munii (pl.)
3rd pl. o-meni si-palimi-si tabu ne-sou la-maisu-giteu mata ne-ria -dratou (tr.), -dra (pl.)
- 118

4.1. Fijian, Polynesian ko before pronouns, the word for “who?”, and names of persons and places appear in most Nakanai dialects as e. In Bileki it is regularly used before personal pronouns (Bileki e-i-au ‘I’, e-rei ‘who?’ and Fijian ko-i-au ‘I’, ko-cei ‘who?’), before proper names in address and optionally in reference, before most place names, before all kinship terms in address and optionally in reference, and before any other word when used as a characterizing name or nickname (magese ‘red’, e-magese ‘the red’ as a name for reddish coloured pigs).

4.2. The pronominal suffix -tou to mark plural or trial (presumably a variant form of tolu ‘three’) is used for the plural in all dialects of Nakanai (see Table 2).

4.3. The pronominal prefixes Fijian ke- and Polynesian ki- are not represented in Nakanai.

4.4. The 2nd person singular possessive suffix -u shared by Rotuman and Polynesian may be represented in Bileki ta-u-me ‘of, to you’, in the series taku ‘of me’, taume ‘of you’, tetala ‘of him,’ -me alone being the usual second person objective suffix, as in ubi ‘to spear’, ubi-me ‘spear you’.

4.5. The possessive classifiers a- and o- peculiar to Rotuman and Polynesian are not found in Nakanai, but the form a- used with possessive suffix in some predicate constructions may be cognate. The form ne- in the Meramera possessives (Table 2) is clearly cognate with the ne, no- of Fijian. Both the a-, o- of Rotuman and Polynesian and the ne-, no- of Fijian, moreover, are reflected in Mangseng in ö-k mla ‘my blood’, i-a mla ‘his blood’, li-k muma ‘my tongue’, i-le pupo ‘its tail’; and in Kapore we find the corresponding kö-k tamudi ‘my viscera’, pu-ka tamudi ‘his viscera’, pu-le kum ‘his heart’. These traits are obviously without diagnostic value for a Central Oceanic subgroup.

4.6. The plural use of the article ne- in Rotuman and Polynesian is not found in Nakanai. Nakanai uses the article la- in much the same manner as its Fijian cognate na-.

4.7. It is evident that Nakanai shares some of the foregoing features with the Central Oceanic languages, but not all of them. It is note-worthy that some of these features are shared by only two of the Central Oceanic trio and not all three, being found in their entirety only in Polynesian. Some, moreover, turn out to have a much wider distribution, appearing not to be Central Oceanic innovations at all. On the basis of these several features, then, Nakanai has about as good a claim to be included in the Central Oceanic group as does Fijian or Rotuman.

5. There is a small list of six vocabulary items which Polynesian shares with Rotuman or Fijian and which Grace regards as innovations peculiar to the Central Oceanic languages. No cognates for any of them appear in Nakanai, except in one instance, and it does not share the metathesis which is the real innovative feature: Fi wasa, wasawasa, Fu vasa ‘sea, ocean’, as compared with Bi parava, ‘strand, shore’, Tk - 119 ma-taw, Gb ma-rawa, ‘sea, ocean’, PAN *sawaN. 11 In this instance there is phonologically no clue as to whether the Fijian form is a borrowing from the Polynesian or not, a fact which tends to make the evidence of this kind inconclusive. Grace did all his work, moreover, with Polynesian as his point of departure, ignoring vocabulary items shared by Fijian and Rotuman but absent in Polynesian. Before we leave this kind of evidence, however, I wish to point to one vocabulary innovation shared by Polynesian, Rotuman, and Nakanai, but not by Fijian.

5.1. The usual word for ‘blood’ in Austronesian languages in the Pacific is derivable from PAN *(dD) aRaq: Fi draa, Tk cca, Gb ra-raa. It appears in 24 of 40 Melanesian languages listed by Codrington. 12 It is also represented in the three Willaumez dialects: Ga a-ndala, Bo dara, Bu ndara, PWm *ndaRa. It appears in the Meramera dialect of Nakanai as dala-gu ‘my blood’. Throughout Polynesia, however, it has been replaced by another word PPN *toto: To, Sm, Ma toto, Ha koko ‘blood’. Rotuman, too, has toto ‘blood’; and significantly, ka-soso is the standard word for ‘blood’ in all Nakanai dialects except Meramera, where the word appears in the meaning ‘red’. Cognates in Fijian are soso-ko ‘viscous’ and soso-ucu ‘mucous’, which match the Bileki word soso ‘to spill out’ in Nakanai. As we shall see, these words are derivable from PAN *cecen or *cencen, but the meaning of ‘blood’ seems to be a special innovation shared only by Polynesian, Rotuman, and Nakanai. At some time in the history of these languages they must have been dialects within a single isogloss boundary.

Examples such as this are hard to find. Fortunately, when we turn to phonological considerations, we get onto much firmer ground.

6. Phonological comparisons between Nakanai and Central Oceanic languages require that we reconstruct the phonology of Proto-Nakanai. This is essential because the phonological picture has been complicated and obscured as a result of the mutual borrowing taking place in Nakanai among geographically contiguous dialects whose speakers interact with each other frequently and intensively, often residing for many years in dialect areas other than those in which they grew up. I cannot review all the evidence here. Table 3 presents a summary of my reconstruction based on the three most distinctive dialects, Bileki, Ubae, and Meramera. Material from other dialects has been used more sparingly because of the uncertain phonemic status of [l] and [r] among them.

- 120
TABLE 3
RECONSTRUCTED PROTO-NAKANAI PHONEMES
Bi Ub Ve Mt Mr TNk PNk PAN
r r r/l r/l l r    
l l r/l r/l l l l r, l, d, D, R
l l r/l r/l n n n n, ñ
l l r/l r/l N N    
g g g g N ng N N
g g g g g g g Ng, Nk
g r g g d d    
g g g g d D d nt, nT, nd, nD, ns, nz, nZ, nc, nj
g r g g g G    
t t t t t t    
t t t t s T t t, T
s s s s s s s s, z, Z, c, j
r s r/l r/l s j    
r r r/l r/l s J j s, z, Z, c, j
k k k k k k    
k k k k q/O K k g, k
O O O O q q q g, k
h h h h O h    
h h h h q H h q; O initial before a
m m m m m m m m
b b b b b b b mb, mp
p p p p p p p b, p
v v v v v v v b, p, w (sometimes)
u- v- v- v- O/o- w- w- w
O O O O O O O h, R (sometimes; O (except initial before a)
a a a a a a a a
e e e e e e e ey, ay
o o o o o, -e o o e, aw
i i i i i i i i, uy
u u u u u, -i u u u, ew, iw, wi (before consonant)

6.1. The decision to consolidate TNk r, l as PNk l; TNk N, ng as PNk N; TNk d, D, G as PNk d; TNk t, T as PNk t; TNk j, J as PNk j; TNk k, K as PNk k; TNk h, H as PNk h was based on evidence that dialect borrowing was responsible for new correspondence patterns.

This evidence included:

(a) The existence of doublets: e.g., Mr o-ate, ate-na, Ub hate, Bi hate ‘liver, internal organ’ along side of Mr Nala-sasaqe, Ub lala-hate, Bi lala-hate ‘to breathe’, and this along side of Mr Nala, Ub gala, Bi gala ‘breath’, giving us the two sets of reconstructed doublets TNk *hate, *HaTe and *Nala, *ngala.

- 121

(b) Two phonemes of a related language or of reconstructed PAN being randomly reflected by two TNk phonemes: e.g., Fijian r, l and PAN r, l are both reflected by TNk r and also by TNk l under no discernibly different sets of conditions. Thus, TNk *rango, Fi lago ‘fly’, PAN *laNaw; Bi *raga (TNk? *raga) ‘jump, start up’, Fi laqa ‘spring, fly up’, PAN *laNkah; TNk *lua, Fi lua ‘vomit’, PAN *luaq; TNk *laNi, Fi lagi ‘sky’, PAN *laNit; Bi rere-tala (TNk? *jere-) ‘to rise (of a star)’, Fi ceere ‘high’; Bi lage (TNk? *lage) ‘to bind on (of fibre rings on stretched ear-lobes)’, PAN *(rR)aNkay. Phonetic representation of the PNk phoneme as [r] or [l] in different dialects, and differential merging with other PNk phonemes would lead in the course of borrowing to the creation of new contrasts, giving rise to the distinct phonemes /l/ and /r/ in the Bileki and Ubae dialects today. I present this example here because the consolidation of TNk l, r as PNk l is admittedly debatable. Further work on the material from the other dialects may require me to take it back. I have no serious reservations about the other consolidations.

6.2. That PN b, p are reflected by both PNk p and PNk v deserves a moment's comment. Both PNk phonemes correspond to Fijian v and Polynesian f. Doublets such as Bi vulo ‘to turn’ and Bi pulo ‘to turn, twist, wring’ tell the story. At some point after PAN b, p had merged, there arose two dialects with the resulting phoneme pronounced [v] in one and [p] in the other. Subsequent borrowing by one from the other was accompanied by phonemic reassignment of the phones in the borrowed forms. This must have happened after the separation of PNk from Early Fijian because the Fijian doublet caa ‘bad’, vu-sa ‘rotten’ (PAN *zaqat) is represented in Nakanai as Bi sa-saha ‘bad, rotten’, pu-sa ‘rotten’.

7. There is a possibility that Fijian and Nakanai share a phonological innovation not known from elsewhere. In Fijian PAN q, h have been lost. But where PAN had an initial a or where the loss of PAN q, h would have resulted in an initial a, Fijian has developed an initial y.

  • PAN *hasaN (? ‘gills’), To aha ‘meaty places’, Fi yasa ‘a side’.
  • PAN *hapir, To ofi ‘near’, Fi yavi-taua ‘to hail a ship from shore’.
  • PAN *aka(DrR), To aka ‘root’, Fi yaka ‘a creeper.’
  • PAN *qatey, To qate, Fi yate ‘liver’.
  • PAN *paqa, To paqapaqa ‘stalk, stem’, Bi vaha, Fi (with metathesis) yava ‘leg’.

In Nakanai PAN q appears as h, PAN h is lost, but PAN initial Oa has been replaced by *ha.

  • PAN *anak (? ‘child’), Bi hala-gu ‘my maternal uncle, sororal nephew or niece’.
  • PAN *apuy, To afi, Bi havi, Ub havi, Mr o-avi ‘fire’, PNk *havi.
  • PAN *abuh, To afu, Ub habu-ro ‘ashes’, Fi yavuyavu ‘ruins after a fire’.

I have no evidence for PAN initial ha, all Nakanai examples having acquired a prefix of some sort. No such development took place before other vowels in initial position either in Fijian or Nakanai.

- 122

Nakanai would seem to explain the source of the mysterious y before initial a in Fijian. Before PAN q was lost in early Fijian it replaced initial O before a. It was subsequently lost in all other positions, except initially before a, where it remained as y. In Pre-Nakanai it likewise replaced initial O before a, but was not subsequently lost in other positions, being preserved as PNk h. This is not likely to have been an independent development.

8. Dempwolff found PAN *c, j, s, z to be represented by Fi s, c and PPN s, h. In order to resolve the problem of a double set of reflexes he posited that Fi s and PPN h reflect PAN *nc, nj, ns, or nz. 13 This solution has been accepted by Dyen, who has added PAN *Z to this series, 14 and by Grace. 15

The same pair of reflexes occurs in PNk as *s, *j. Taken together they correspond regularly with Fi s, c and PPN s, h. Taken separately, however, the correspondences are highly irregular. This we might expect if the two phonemes in each case were morphophonemic alternates with and without prenasalization. But there are some additional facts worth noting.

8.1. Correspondences between PPN *s, *h and Fijian s, c are completely random. Thus Fijian and Polynesian cognates for PNk *s and *j show the following correspondences when compared with one another. 16

Fi c and PPN *h 8 instances
Fi c and PPN *s 6 instances
Fi s and PPN *h 6 instances
Fi s and PPN *s 7 instances

When we compare PNk with Fijian and with PPN the situation is less completely random.

PNk *s and Fi s 15
PNk *s and Fi c 9
PNk *j and Fi s 4
PNk *j and Fi c 11
PNk *s and PPN *s 12
PNk *s and PPN *h 6
PNk *j and PPN *s 33
PNk *j and PPN *h 11

The latter comparisons strongly suggest the onetime equations: (1) PNk s, Fi s, PPN s; and (2) PNk j, Fi c, PPN h.

Our equations come out differently from those accepted for Fijian and Polynesian. This reveals the essential weakness of Dempwolff's explanation. The material suggests that in Central Oceanic there were two phonemes representing PAN c, j, s, z, Z. They were variously pronounced in neighbouring dialects with the allophonic range for one in one dialect overlapping the allophonic range for the other in the next dialect. Subsequent borrowing between dialects led to the equating of what was one phoneme in the donor language with the other one in the - 123 borrowing language. This process produced the doublets which we find in all three language groups. As this process continued, correspondences between the two phonemes became more and more scrambled, approaching almost complete randomness between Fijian and Polynesian, which remained in proximity with one another longer. The lesser degree of randomness of correspondence between Nakanai and Fijian and between Nakanai and Polynesian reflects the geographical separation of Proto-Nakanai from Early Fijian and Proto-Polynesian, a fact which arrested the scrambling process in interaction with these dialects before it had gone so far. It continued among the Nakanai dialects themselves, of course, as revealed by TNk j, J (Table 3), but we have allowed for this in our reconstruction of PNk.

8.2. The occurrence of some doublets and unexpected correspondences in Nakanai, moreover, suggest that PAN nc, nj, ns, nz, nZ are reflected by PNk d and not PNk j or s. In this respect Nakanai is like Sa'a, where PAN c, s, j, z, Z are reflected by t (s before i and u), while PAN nc, ns, nj, nz, nZ are reflected by Sa'a d. 17

  • PAN *Zalan, Fi sala, Sa tala ‘path’: Mr sala ‘outdoors’, PNk *(sj)ala; PAN *nZalan: Bi gala, Ub rara (rala) ‘outdoors’, PNk *dala. See also Ga dala, Bo dala, Bu ndala ‘path’, PWm *ndala, when PAN Z is normally reflected by PWm r. Compare also Fi (Nandronga dialect) sa-levu ‘road, path’ with Bi ga-uru ‘road, path’, in which the -levu and -uru, both mean ‘big, main’.
  • PAN *quzuN, Fi ucu ‘cape, mountain peak’, Ro qusu-qi ‘head’; PAN *qunzuN: Bi hugu, Ub huru, Mr o-udi ‘carry on the head’, PNk *hudu.

8.3. In summary, it appears that PNk s, j correspond to Fi s, c and PPN s, h. Their mutual correspondences, moreover, suggest that at one time in their history these three language groups were three geographically contiguous dialects. Mutual borrowing led to a progressive scrambling of the original paired correspondences. This scrambling can no longer be attributed to morphophonemic alternation between nasal and oral grades, as they have been called, for the nasal grade is reflected by PNk d. Again Nakanai shows its close relation to Central Oceanic languages and clarifies the interpretation of their phonology.

9. According to the evidence compiled by Dempwolff, 18 PAN c fell together with PAN j, s, z, Z, which are reflected in Fijian by s, c, in PPN by s, h, and in Sa'a by t (s before i, u). Dempwolff's material shows PAN c well represented in Fijian, in six instances as Fi s and in four instances as Fi c. For the other languages the evidence is meagre: PPN s (two instances), PPN h (none); Sa t (two instances), Sa'a s (none). 19

- 124

9.1. Such evidence as there is for Nakanai indicates that PAN c fell together here also with PAN j, s, z, Z as PNk s, j.

  • PAN *cacaq, Fi caacaa ‘potsherds, broken pieces’, Sa tata ‘to scatter’; Bi sara ‘small cut Nassa Calossa shells used for decoration’, PNk *sa(j)a.
  • PAN *cabut, Fi cavu, cavut-a ‘to uproot, eradicate’, Sa tahu'-i ‘to take to pieces’: Bi savu ‘to kill, destroy’.

PAN *nc fell together with PAN nj, ns, nZ, nt, nT, nd, nD as PNk d.

  • PAN *nceguk (Dempwolff *ceguk): Bi do-doki ‘to hiccup’ (presumably a loan from Meramera), from which we infer PNk *doku.
  • PAN *camuk, To hamu ‘to eat of one food only’: PAN *ncamuk, Sa ma-damu ‘to smack’ (Dempwolff), Ul dämu ‘to chew betel’, i-demu ‘lime spatula’: Bi damu ‘lime spatula’ (presumably a loan from Meramera), from which we infer PNk *damu.

9.2. Grace states that in Rotuman PAN c also fell together with PAN j, s, z, Z, as Ro s. 20 He gives no examples, however, apparently assuming that the pattern for Rotuman was the same as for Fijian and Polynesian. I have found three instances for Rotuman, and in all of them PAN c is Ro t.

  • PAN *ceguk: Ro toqo ‘to belch’.
  • PAN *cencen or *cecen: Ro toto ‘blood’ (see above, para. 5.1), Fi soso-ucu ‘nasal mucous’, PNk *ka-soso ‘blood’.
  • PAN *cabut: Ro tau ‘to pick, pluck’, Fi cavu, cavut-a ‘to uproot, pluck’, Bi savu ‘to kill, destroy’.

Rotuman is thus like Trukese and Gilbertese, where PAN j, s, z, Z appear as Tk t and Gb r, but PAN c appears as Tk s and Gb t.

  • PAN *(s)akay: Tk tää-n ‘rising of’, Gb rake ‘up, east’.
  • PAN *sawaN: Tk ma-taw, Gb ma-rawa ‘sea’.
  • PAN *sakit: Tk me-tek, Gb ma-raki ‘to hurt, pain’.
  • PAN *zaqat: Tk ja-taa-j ‘destroy, damage it’, Gb raa ‘bad, ruined’.
  • PAN *cirit: Tk siir ‘urine’, Gb tii ‘to spurt’.
  • PAN *ceguk: Tk ssyk ‘to hiccup’ (from *sosoku).
  • PAN *cabut: Gb tau-a ‘to sieze, grasp’.

9.3. We may diagram the foregoing evidence as follows:

PAN s z Z j c
PNk j/s j/s j/s j/s j/s
Fi c/s c/s c/s c/s c/s
PPN h/s h/s h/s h/s h/s
Ro s s s s t
Tk t t t t s
Gb r r r r t

This would make it appear that if there is justification for regarding Polynesian, Fijian, and Rotuman as constituting a very closely related Central Oceanic group, we must also include Nakanai within it. Furthermore, we must conclude that Nakanai is more closely related to Fijian than Rotuman is. - 125 9.4. There is an alternative explanation, however, which we must consider. I have found one instance each for Polynesian and Fijian where PAN c appears unexpectedly as t.

  • PAN cemeD: Fi tomo ‘dirty, untidy (of nose, face)’ beside somo ‘stained black (of sinnet)’, somo-ta ‘to stain black’.
  • PAN *cencen: Fi soso-ko ‘viscous’, soso-ucu 'nasal mucous (Dempwolff); To, Sm toto, Ha koko ‘blood’, PPN *toto; Ro toto ‘blood’; Bi soso ‘to spill out’, Bi, Ub ka-soso ‘blood’, Mr ka-soso ‘red’, PNk *ka-soso.

We can interpret Fi tomo and PPN toto as the result of borrowing, perhaps from Rotuman, where PAN c appears as t.

On the other hand we might interpret PPN toto as reflecting PAN *nc. On this basis we might also argue that Ro t reflects not PAN c but PAN nc. In support of this we have the example of Nk dodoki ‘to hiccup’, which I have derived from PAN *nceguk. But we cannot so easily interpret Fijian tomo in this way, nor can we readily derive Tk s or Gb t from PAN nc, for the other prenazalized consonants with which PAN nc would presumably have merged are reflected by Fi d, Tk c, and Gb r.

We come back, therefore, to our first conclusion; Nakanai, Fijian, and Polynesian go together in their handling of PAN c, whereas Rotuman goes with the Micronesian languages. Phonologically, Nakanai fits better into Grace's Central Oceanic group than Rotuman does.

10. The lexicostatistical evidence, for what it is worth, thoroughly reinforces this conclusion. For the 200 word list proposed by Swadesh, Grace gets the following percentages of cognates on two separate calculations: 21

Calculation 1 Calculation 2
Maori-Tongan 38% Maori-Tongan 40%
Fijian-Rotuman 22% Fijian-Rotuman 23%
Maori-Rotuman 21% Maori-Rotuman 23%
Maori-Fijian 21% Maori-Fijian 23%
Fijian-Mota 20% Fijian-Mota 23%
Maori-Mota 18% Maori-Mota 21%
Rotuman-Mota 15% Rotuman-Mota 19%

Compare this with our percentages, using essentially the same list, where we have 28% cognates for Bileki-Fijian (para. 3.1.). The percentages for Bileki-Fijian are sufficiently higher than for Fijian-Rotuman, Maori-Rotuman, and Maori-Fijian that, even allowing for error, Nakanai must be classed as a Central Oceanic language.

11. If Nakanai is Central Oceanic, we can argue that it represents a group of Central Oceanic speakers who were left behind in the migrations into the Central Pacific area. Or we can argue that Nakanai results from a back-migration from the outer islands westward into Melanesia. The latter interpretation is the one which can be more - 126 readily supported. The “principle of least moves” requires that we bring Nakanai from the Central Pacific. The Polynesian outliers provide a precedent, and Sharp's analysis of Pacific voyaging indicates that the probabilities favour east-to-west population movements.

Nakanai provides evidence that the process of back-migration from the Central Pacific involved other languages as well as Polynesian ones. If this process has been going on ever since the Central Pacific was first settled, it would help account for some of the puzzling complexities which characterize linguistic relationships in Melanesia.

REFERENCES
  • CODRINGTON, R. H., 1885. The Melanesian Languages, Oxford.
  • DYEN, Isidore, 1951. “Proto-Malayo-Polynesian *Z.” Language, 27:534-40.
  • — — 1953. The Proto-Malayo-Polynesian Laryngeals. Baltimore.
  • DEMPWOLFF, Otto, 1934-8. Vergleichende Lautlehre des austronesischen Wortschatzes, 3 vols., Beihefte zur Zeitschrift für Eingeborenen-Sprachen, 15, 17, 19.
  • GRACE, G. W., 1955. “Subgroupings of Malayo-Polynesian: A Report of Tentative Findings.” American Anthropologist, 57:137-39.
  • — — 1959. The Position of the Polynesian Languages within the Austronesian (Malayo-Polynesian) Language Family. Indiana University Publication in Anthropology and Linguistics and Memoir 16 of the International Journal of American Linguistics, Baltimore.
  • GUDSCHINSKY, Sarah, 1956. “The ABC's of Lexicostatistics (Glottochronology.” Word, 12:175-210.
  • HYMES, D. H., 1960. “Lexicostatistics So Far.” Current Anthropology, 1:3-44.
  • SWADESH, Morris, 1950. “Salish Internal Relationships.” International Journal of American Linguistics, 16:157-67.
1   Grace 1959:65.
2   Since this was written, Dyen (1960:180-4) has shown the inadequacy of Grace's evidence for the thesis that these three languages form a sub-group as compared with any other Austronesian language. Dyen's arguments do not directly support my thesis about the position of Nakanai, but they provide additional support for my contention that Grace has not made a case which would exclude Nakanai from the Central Oceanic group.
3   Grace, 1955:338-9.
4   Information about Nakanai dialects and neighbouring languauges comes from data which Ann Chowning, C. A. Valentine, and I collected in New Britain in 1954-6.
5   Sharp 1956.
6   Swadesh 1950; Gudschinsky 1956; Hymes 1960.
7   In his massive lexicostatistical survey of Austronesian languages, Dyen finds 20.7% to be the highest percentage of shared cognates between West Nakanai (Bileki) and any language in his so-called Oceanic sub-group, which includes Fijian (as reported by him in a paper read before the Society of Asian Studies, April 11, 1960). This discrepancy between his figures and mine may reflect a somewhat wider recognition of cognates on my part as a result of my efforts to reconstruct Proto-Nakanai; it may also reflect less rigour in my application of lexicostatistical techniques. Dyen's use of an electronic computer to make the lexicostatistical counts, on the other hand, may have led to programming procedures which systematically produced lower percentages. In any event, my own methods were consistently applied, so that the differences shown here should be reliable within a relative if not an absolute chronology.
8   In the orthographies used here, D in Mangseng represents the sound of English “th” in “they”; N represents the sound of English “ng” in “sing”; x represents a voiced dorsal spirant; q represents the glottal stop; and ä and ö represent open e and o. For Fijian and Sa'a, however, I have followed the standard orthography of dictionaries, and for PAN I employ the orthography used by Dyen (1953).
9   Grace 1955.
10   Grace 1959.
11   The abbreviations for languages are: Bi—Bileki (Nk), Bo—Bola (Wm), Bu—Bulu (Wm), Fi—Fiji, Fu—Futuna (PN), Ga—Garua (Wm), Gb—Gilbert Islands, Gk—Gaikeke (Nk), Ha—Hawaii (PN), Kp—Kapore, KS—Kai-Sisimi (Nk), Ma—Maori (PN), Mr—Meramera (Nk), Ms—Mangseng, Mt—Maututu (Nk), Nk—Nakanai, PAN—Proto-Austronesian, PN—Polynesian, PNk—Proto-Nakanai, PPN—Proto-Polynesian, PWm—Proto-Willaumez, Ro—Rotuma, Sa—Sa'a, Sm—Samoa (PN), Ta—Tarobi (Nk), Tk—Truk, TNk—Tentative Proto-Nakanai, To—Tonga (PN), Ub—Ubae (Nk), Ul—Ulawa, Ve—Vele (Nk, including Gk, KS, and Ta), Wm—Willaumez.
12   Codrington 1885.
13   Dempwolff 1934-8, Vol. 2:138, 192.
14   Dyen 1951.
15   Grace 1959:27.
16   Taken from materials in Dempwolff 1934-8, Vol. 3.
17   Dempwolff 1934-8, Vol. 2:156.
18   Dempwolff 1934-8, Vol. 2:166, 192.
19   Dempwolff 1934-8, Vol. 3.
20   Grace 1959:29-30, 70.
21   Grace 1959:68.