Volume 76 1967 > Volume 76, No. 3 > The relationships of Polynesian Outlier languages, by Andrew Pawley, p 257 - 296
- 258 Page is blank- 259
THE RELATIONSHIPS OF POLYNESIAN OUTLIER LANGUAGES 1
0. Summary. It is argued here that all 12 Polynesian “Outlier” languages for which data are available belong to the Nuclear Polynesian subgroup, which includes all well-described Polynesian languages apart from Tongan and Niuean. It is further argued that Nuclear Polynesian divides into an Eastern Polynesian subgroup on the one hand, and a coordinate subgroup including Samoan, East Futunan, the Ellice Is. dialects of Vatipu and Nanumea, Tokelauan, Pukapukan, and probably all known Outliers, - 260 on the other. The name Samoic-Outlier (SO) is suggested for this last subgroup.
Within SO at least three subgroups appear to be quite well marked; one comprises the two north-western Outliers, Nukuoro and Kapinga-marangi; a second consists of the central Outliers, Sikaiana, Takuu, and Laungiua (Ontong Java). A third consists of two New Hebrides Outliers, Mele-Fila and Futuna-Aniwa. Apart from dialect clusters (such as exists in the Ellice Is.) no other clearly marked subgroups have been identified, but the restricted distribution of certain lexical and grammatical features provide evidence either of subrelationship or borrowing in several further cases.
1. Introduction. The 30 or so Polynesian (PN) languages, 2 distributed across the Pacific from 155° E. to 109° W. longitude, and from 22° N. to 48° S. latitude, are sometimes divided, for convenience of reference, into two geographic categories: “Triangle” and “Outlier” languages. The apices of the “Polynesian Triangle” are Hawaii in the north, New Zealand in the south, and Easter Island to the east. Within this vast area some 16 or more distinct languages—all of them PN—were spoken at time of first European contact.
Further west, within the much smaller area between 4° N. and 20° S. latitude, and 155° E. and 170° E. longitude, lie the Polynesian Outliers. There are some 14 or 15 Outlier languages, spoken by small communities occupying islands inside or on the fringes of geographic eastern Melanesia and Micronesia; this is a linguistically diverse area where perhaps a hundred non-PN languages are spoken.
Few of the Outliers are well documented. Fairly extensive data, including detailed phrase grammars, and some lexical and text materials are available for four of them: Nukuoro (NUK), Kaingamarangi (KAP), Sikaiana (SIK), and Luangiua (LUA) (Ontong Java). Six others (Mae (MAE), Mele-Fila (FIL), West Uvea (WUV), Futuna-Aniwa (WFU), Rennell-Bellona (REN), and Tikopia (TIK) are known, in lesser detail, from grammatical sketches and/or texts and wordlists and very meagre grammatical notes. 3 Published data for Anuta (ANA), Takuu (TAK), Pileni (PIL), Taumako (TAU), Nukuria (NRA), and Nukumanu (NMU) are nil or almost nil; however, observers have stated that TAU and PIL are perhaps dialects of one language, while NRA and NMU are very- i - ii - 261
The weight of the evidence so far advanced suggests that the languages of Triangle Polynesia fall into two (coordinate) major subgroups: Tongic and Nuclear Polynesia. 6 Tongic (TO) has only two certain members: Tongan and Nuiean. With the possible exception of East Uvean, 7 all other well-described Triangle languages belong to the Nuclear Polynesian (NP) subgroup.
Fairly substantial evidence supports the division of NP into two coordinate subgroups: Eastern Polynesian (EP) and Samoic (SM). 8 EP includes New Zealand Maori (MAO), Rarotongan (RAR), Tahitian (TAH), Hawaiian (HAW), Tuamotuan (TUA), N.W. and S.E. Marquesan (MQA), Mangarevan (MVA), and Easter Island (EAS). It has been argued that SM includes Samoan (SAM), East Futuna (EFU), and Ellice (ECE) (represented by the Vaitupu dialect). Tokelauan (TOK), Nanumea (NAN), and Pukapukan (PUK), 9 must now be added to SM. Within EP, a further subgrouping has been suggested by Elbert and Green, both of whom regard EAS as the sole member of a subgroup coordinate with the remaining EP languages. 10
While the relationship of the Outliers to the Triangle languages has been in dispute for half a century, there have been few serious investigations of the problem. There are two schools of thought on Outlier origins. Most writers have regarded the Outliers as deriving from languages spoken in Triangle Polynesia, or from other Outliers which themselves derived from Triangle languages. Elbert in his 1953 paper classified KAP and TIK as NP, and included TIK in his Samoic subgroup. 11 In an earlier paper, I noted that Outliers share some of the morphological - 262 innovations characterising NP languages, but excluded all but TIK from classification. 12
However, the most detailed comparative treatment of Outlier languages to date is that of Bayard, in a recent unpublished study. 13 He considers phonology, some morphological features, basic vocabulary and kinship terminology, in addition to non-linguistic data, in an attempt to determine the sources of each Outlier culture. Bayard includes all Outliers in NP. While he does not explicitly carry his linguistic subgrouping any further than this, consideration of ethnographic as well as linguistic evidence allows him to reach several interesting conclusions about the origin of particular Outlier cultures. Worthy of note among his conclusions is the importance attributed to East Futuna (Hoorn Is.) and the Ellice Is. as sources of both primary and secondary settlements. Also noteworthy is his claim that most Outlier cultures have multiple origins, with primary and secondary settlement from several sources in Triangle and/or Outlier Polynesia, and/or Melanesia and Micronesia.
The linguistic results of secondary settlement and other significant contracts posited are not defined by Bayard. Elbert, however, has shown that the unusual phonological features of certain Outliers can be directly attributed to the influence of nearby non-PN languages. 14 In fact, borrowings from non-PN languages are usually easily identified; the chief problem is in identifying borrowings from other PN languages.
A different viewpoint concerning Outlier origins has been expressed by Capell, who argues that the grammatical core of certain Outliers, notably PIL, MAE, WFU, and FIL, derives “from an earlier Polynesian language left on the western islands as the ancestors moved eastwards”. 15 While admitting that “later influences from the east have contributed much vocabulary to the outliers”, he considers that this has not “been sufficient to alter the grammatical basis of the earlier Polynesian language. . . . There is so much in [WFU], Mae, Pilheni and other languages that has no parallel further east”. 16 He has, however, said very little about the evidence which might support this thesis. 17 I know of no phonological and only one morphological feature which might suggest that the languages of the Triangle form a subgroup apart from certain Outliers (this being the subgrouping implicit in Capell's remarks). 18- 263
Much new information has come to hand over the last two or three years on both Outlier and Triangle languages. Consideration of these data provides evidence supporting Bayard's conclusion that all Outliers belong to the NP group, and suggests the following subclassification of NP languages:
NP divides into an Eastern Polynesian (EP) and a Samoic-Outlier (SO) subgroup. EP members are as stated earlier. SO is an expansion of the SM group outlined above. In addition to the Triangle languages SAM, TOK, PUK, EFU, VAI, and NAN, it includes the following 12 Outliers: SIK, LUA, TAK, TIK, NUK, KAP, REN, PIL, MAE, WUV, WFU, and FIL.
Within SO subrelationships are less well-defined. This is in part because (as Bayard noted) many SO communities, speaking languages which are to some degree mutually intelligible, have remained in constant or occasional contact, with considerable borrowing resulting in at least some cases. 19 However, at least three subgroups appear to be quite well marked. One comprises the northwestern Outliers, NUK and KAP; another includes the central Outliers, SIK, TAK, and LUA. A third consists of two New Hebrides Outliers, Mele-Fila and Futuna-Aniwa. The two Ellice Is. communities VAI and NAN, while showing some differences, can be considered to speak dialects of a single language (although this similarity may result from continuing contact between Ellice communities over a long period, rather than from very recent separation). 20
No other subgroups of SO are presently obvious. However, some further evidence of either subrelationship or extensive borrowing is - 264 provided by the distribution among SO members of small clusters of uniquely shared lexical, morphological and morphophonemic features. For example, EFU shares a few probable innovations with TIK, WFU, WUV, MAE, and FIL, and VAI and NAN share a few probable innovations with SIK, TAK, LUA, KAP, and NUK. In neither case is the quantity or quality of evidence sufficient to firmly establish a subgroup. But it does seem likely that better descriptive data will yield clear evidence both of further subgroups and of the directions of later borrowing between Outlier and Triangle languages.
At present SAM, REN, PUK, TOK, and EUV, among SO languages, stand without any clearly marked relationships. This may change when fuller data are available; on the other hand, this situation may reflect early divergence of each of these languages from the rest of SO.
The remaining body of the paper is organised as follows: Sec. 2 contains a brief summary of the evidence for the membership of the known Outliers in NP. Sec. 3 contains the evidence for the SO subgroup of NP. Sec. 4 contains discussion of the internal relationships of SO languages.
2. The Outliers and Nuclear Polynesian
Some phonological and morphological evidence for the NP and TO subgroups has been stated elsewhere, 21 and will not be repeated in detail here. With the exception of three further possible morphophonemic innovations noted recently 22 and some revisions discussed under (2) below, I outline only briefly those innovations cited previously as characterising NP. However, the replacements of PNP features in the Outliers, and in PUK, NAN, TOK, and certain other Triangle languages are cited here in full for the first time.
TABLE I: PPN phonemes and their correspondences in daughter languages. 23
All NP members have merged PPN *l and *r (usually either as /l/ or as /r/ and have lost PPN *h. Both members of TO, in contrast, retain the distinction between PPN *l and *r (*l is reflected by /l/, *r by zero), and reflects both *s and *h as /h/; moreover, in the TO languages, PPN *a has sometimes become /e/ near /i/ or /e/, and *a has sometimes become /o/ near /o/ or /u/.
The following innovations are attributed to Proto-Nuclear Polynesian (PNP):
(2) PPN *eni˜*ani “this, these” and *ena˜*ana “that, those (near addressee)” are reconstructed from TON, EUV eni, NIU nai˜nei˜hanai˜hanei “this, these”, TON, EUV ena, NIU na˜hanaa “that, those (near addressee)”, plus the NP demonstratives cited below, and the forms for other Austronesian languages cited in discussion of innovation (8). These PPN (pronominal) demonstratives are reconstructed as occurring alone in the nucleus (e.g., as *ko eni “this one, these ones”), and postponed to the nucleus obligatorily paired with the definite article *(t)e (e.g., as *(t)e fale eni “this house/these houses”). The demonstratives have this distribution in all known PN languages, and in fact in most Eastern Oceanic (EO) languages. NP languages, however, exhibit certain innovations in the form and structure of the demonstratives and most have extended the distribution to permit optional occurrence preposed to the nucleus (as e.g., SAM ?o lenei fale; MAO ko teenei fale “this house”).
The structural features characteristic of NP pronominal demonstratives are (a) their obvious internal structure, consisting of an article plus a marker of position relative to speaker and/or addressee, and (b) the distinction between singular and plural number marked in the article member of the construction. The TON and EUV forms eni, ena, ee, may optionally be analysed as consisting of the definite article e- plus a position marker -ni, -na, or -e, but as none of the other nominal articles combine in this way with a position marker, this bimorphemic analysis is not strongly required. A somewhat stronger case can be made for analysing the NIU demonstratives hanai, etc., but again the pattern is not productive. Secondly, and related to this invariability of form, the TON, EUV, and NIU demonstratives do not distinguish number, TON ko eni, for example, can mean either “this (one)” or “these (ones)”. 24 In fact, as is noted in the discussion of feature (6), TON and NIU do not distinguish number in any of their several nominal articles.
Besides these structural and semantic features, the formal features characterising NP demonstratives apart from those of TO and EUV consist of the form nei (presumably by metathesis of PPN *eni, which has also occurred in NIU) and of the long vowel in NP naa. The PNP demonstratives are reconstructed as follows (where . . . represents the phrase nucleus): (a) *teenei . . .˜ te . . . nei ˜ te teenei “this”; (b) *teenaa . . . ˜ te . . . naa te . . . teenaa “that (near addressee)”; (c) *(te-)*eng;a . . . nei ˜ (ee)nei . . . “these”; (d) (te-)ŋa . . . naa ˜ (ee)naa . . . “those (near addressee)”. As can be seen from the reflexes cited below, most NP languages have two or three alterant forms for each demonstrative - 267 meaning, e.g., SAM leenei fale ˜ la fale nei ˜ le fale leenei “this house”.
(a) PNP * teenei. . . ˜ te . . . nei ˜ te . . . teenei “this” is replaced by: WUV de . . . nei; WFU ta . . . tenei ˜ ta . . . nei; FIL te . . . ne ˜ te . . . nane; MAE re . . . ni ˜ -nii; NUK te . . . nei; KAP ti . . . nei W ti. . . teenei; REN, TOK teenei. . . ˜ te . . . nei; SIK te . . . nei ˜ teenei. . .; LUA . . . keeŋei ˜ ke . . . ŋei; PUK te . . . teenei; MAO, RAR teenei . . . ˜ te . . . nei; EAS te . . . nei; SAM leenei ˜ le . . . nei ˜ le . . . leenei; NAN, VAI te . . . nei ˜ te . . . teenei.
(b) PNP *teenaa . . . ˜ te . . . naa ˜ te . . . teenaa “that (near addressee)” is replaced by: WUV de . . .naa; FIL te . . . ŋa ˜ te . . . ŋana; MAE . . . ŋa ˜-naa; NUK te . . . na; KAP ti . . . naa ˜ ti. . . teenaa; SIK te . . . naa ˜ teenaa; LUA keeŋaa ˜ ke . . . ŋaa; REN, TOK teenaa . . . ˜ te . . . na; MAO, RAR teenaa . . . ˜ te . . . naa; HAW keenaa . . . ˜ ke . . . naa; EAS te . . . ena; PUK, VAI, NAN te . . . naa ˜ te . . . teenaa; SAM leenaa . . . ˜ te . . . naa W le . . . leenaa.
(c) PNP *(ee)nei te-na . . . nei “these” is replaced by: WUV Ø . . . naa (where Ø = zero in the article slot): WFU ŋa . . . aŋanaa 25; FIL a . . . ne ˜ a . . . ŋane; NUK a-nei . . . ˜ te-ŋa . . . nei; KAP na . . . nei ˜ a-nei . . .; MAE a-nii; 26 SIK na . . . teenei ˜ na . . . nei; LUA ŋaa . . . nei; TOK . . . leenei ˜ naa . . . ieenei; MAO eenei . . . ˜ na . . . nei; RAR eenei. . . ˜ ŋaa . . . nei ˜ te . . . ŋaa . . . nei; EAS na . . . nei; EFU, VAI Ø . . . nei; SAM Ø . . . nei ˜ Ø-nei. . .; PUK na . . . nei; NAN te-ŋa . . . nei ˜ Ø . . . nei.
(d) PNP *(ee)naa ˜ te-ŋa . . . naa “those (near addressee)” are replaced by: WUV Ø . . . naa; WFU ŋa . . . aŋana; FIL a . . . ŋana; NUK a-na . . . ˜ te-ŋa . . . na; KAP na . . . naa ˜ a-na .. . . .;SIK na . . . teenaa; LUA ŋa . . . ŋaa; MAE a-naa; MAO eenaa . . . ˜ ŋaa . . . naa; RAR eenaa . . . ˜ ŋaa . . . naa ˜ te-ŋaa . . . naa; EAS ŋa . . .ena; SAM Ø . . . naa ˜ Ø-naa . . .; VAI, NAN Ø . . . naa; PUK na . . . naa; TOK . . . ieenaa ˜ naa . . . ieenaa.
(3) PPN *ni “near speaker, now”, a postnuclear marker in nominal and verbal phrases, is replaced by PNP *nei: KAP, NUK, SIK, TIK, PIL, REN, WUV, WFU, TAK, NAN, PUK, MAO, SAM nei; LUA ŋei; FIL ne ˜ nei; MAE nii ˜ ni “near speaker”. Besides the formal innovation there is a possible distributional change: the reflex of nei occurs postponed to the nucleus in all NP languages, but in at least two languages, SAM and MAO, it also occurs preposed to the nucleus following possessive pronouns, as SAM lo ?u nei fale, MAO tooku nei fare “my (very own) house”. This suggests that *nei had this wider distribution in PNP, though confirming evidence is required from other NP languages.
(4) PPN *ha “indefinite article” is replaced by PNP *se “indefinite article singular” and *ni “indefinite article plural”.- 268
PNP *se “indefinite article, singular” is replaced by: KAP, NAN, LUV, EUV, WUV he, SAM, VAI, EFU, TIK, NUK se, WFU sa ˜ se “indefinite article singular”, and by MAE se, TOK, MAO, HAW, TUA he, RAR ?e “indefinite article”.
PNP *ni “indefinite article, plural” is replaced by: KAP, NUK, VAI, NAN, PIL, TIK, SIK, REN, PUK, SAM ni, LUA ŋii, EFU niiki ˜ ni-“indefinite article, singular”. Note also EUV hina “indefinite article plural”. TON ni?ihi “several, some” (a numeral-class base).
(5) PPN *taha “one, a certain one, another one” (a somewhat tentative reconstruction) is replaced in PNP by *tasi (the regular reflex would be *taa). *tasi is reflected by: SIK, KAP, NAN, TOK, WUV, MAO, EAS, MQA tahi, TIK, REN, EFU, WFU, SAM tasi, HAW kaahi, LUA kahi, RAR, MVA ta?i, PIL tai, as a base meaning “one”. PNP *tasi is also reconstructed as occurring, in sequence with an article, in the prenuclear demonstrative/position-marker slot in nominal phrases, carrying the meaning “a certain one, another one”. It is reflected in this function by SAM, EFU, EUV, REN, WFU, MAO, RAR, MQA, TOK, EAS (information is lacking for most other languages), as MAO tee-tahi mea, SAM le-tasi mea “a certain thing, another thing”.
(6) The distinction made by most NP languages between singular and plural number in the article slot in demonstratives and preposed possessive pronouns (as well as articles alone) in nominal phrases, is regarded as reflecting a structural innovation of PNP. No comparable distinction is made in TON or NIU, or in other Eastern Oceanic languages. 27 The only NP languages which apparently do not make this distinction in all the markers in question are LUA (which has lost the preposed dual and plural possessive pronouns, though it distinguishes number in the singular possessive pronouns, articles, and demonstratives) and FIL (which possibly does not distinguish singular and plural demonstratives, though it has the distinction elsewhere).
Recently two other probable morphophonemic innovations of PNP have been noted. These are given extended discussion below.
(7) PPN dual and plural person-markers (except second person) in the nucleus slot are reconstructed as containing a phonemically short (single) *a but are replaced by PNP forms which have phonemically long (geminate) *aa, as follows:
The short *a in the PPN forms is reconstructed on the basis of the TON and NIU reflexes, and cognates in closely related non-PN languages. - 269 All NP languages for which vowel length data are available show long /aa/ in at least some replacements of the PPN person-markers, and in fact most have long /aa/ in all cases (see forms cited below). In such instances of disagreement between TO and NP, evidence from non-PN languages sometimes permits unambiguous reconstruction of the PPN forms. Unfortunately, reliable information about the phonemic status of vowel length is available for very few Oceanic languages outside of PN. Some of the “dialects” of Fijian, 28 e.g., Bauan and Nadrogaa, distinguish long and short vowels in certain positions, and both exhibit short /a/ in the nuclear pronominal forms cognate with the TON and NIU person-markers, as follows:
Cognates in Nguna, a New Hebrides language which has a contrast between short and long vowels, also show short /a/, 29 and it appears probable that this is also the case for certain closely related languages of the Southeast Solomon Is. 30 The fact that the only known cases of long /aa/ in the 1st inc., 1st exc, and 3rd pers. non-singular person-markers are found in NP languages requires the tentative conclusion that this feature was innovated by PNP. More definite conclusions await the appearance of phonemically accurate descriptions of other Eastern Oceanic languages, and particularly, of the Outliers TIK, PIL, FIL, WFU, ANU, MAE, and WUV, and of EFU, EUV, and Niuafo'ou. In the forms cited below capitalised A marks a vowel of uncertain phonemic length.
PNP *kitaaua “1st pers. inc. dual, nuclear” is reconstructed from the following: REN kitaaua, SAM ?itaa?ua ˜ taa?ua, SIK, MAO, RAR, NAN, VAI, PUK, TOK taaua, HAW, LUA kaaua, KAP kitaua, NUK kitau, WFU a-kitAua, MAE, FIL, TIK tAua, WUV gitAua, PIL kitAua.
PNP *kimaaua “1st pers. exc. dual, nuclear” is reconstructed from- 270
the following: REN kimaaua, SAM ?imaa?ua ˜ maa?ua, SIK, MAO, RAR, NAN, VAI, PUK, TOK, HAW, LUA maaua, KAP kimaua, NUK kimau, MAE, FIL, TIK, mAua, WUV gimAua, PIL kimAua.
PNP *kilaaua “3rd pers. dual, nuclear” is reconstructed from REN kigaaua, SAM ?ilaa ?ua ˜ laa ?ua, MAO, RAR raaua, SIK, NAN, VAI, LUA, HAW, PUK, TOK laaua, KAP kinaua ˜ meamaa, NUK kilau, MAE, FIL, TIK rAua, WUV gilAua, PIL kilAua.
PNP *kitaatou “1st pers. inc. pl., nuclear” is reconstructed REN kitatou, SAM ?itaatou ˜ taatou, SIK, MAO, RAR, EAS, NAN, VAI, PUK, TOK taatou, HAW, LUA kaakou, NAP kitaatou, NUK kitateu, MAE, FIL, TIK tAtou, WUV gitAtou, PIL kitAtou; note also WFU a-kitAtou “1st pers. inc. trial, nuclear”.
PNP *kimaatou “1st pers. exc. pl., nuclear” is reconstructed from: REN kimatou, SAM ?imaatou ˜ maatou, SIK, MAO, RAR, EAS, NAN, VAI, PUK, TOK maatou, HAW, LUA maakou, KAP kimaatou, NUK kimateu, MAE, FIL, TIK mAtou, WUV gimAtou, PIL kimAtou; note also WFU a-kimAtau “1st pers. exc. trial, nuclear”.
PNP *kilaatou “3rd pers. pl., nuclear” is reconstructed from: REN kigatou, SAM ?ilaatou ˜ laatou, SIK, NAN, VAI, PUK, TOK laatou, MAO, RAR raatou, HAW, LUA laakou, KAP kinaatou, NUK kilateu, MAE, FIL rAtou, WUV gilAtou ˜ gilehea, PIL kilAtou; note also WFU a-kirAtou “3rd pers. trial, nuclear”.
Among those NP languages where vowel quantity is known, the only cases of short /a/ occur in REN, KAP, and NUK (in the KAP dual forms, in the REN plural forms, and in both dual and plural forms in NUK). However, in phrase final position following ni- “possessive marker” NUK has the following: -taau, -maau, -laau, -taateu, -maateu, -laateu, respectively; moreover, the KAP dual person-markers have long /aa/ in the preposed possessive slot. There is some evidence that KAP and NUK form a subgroup, so that their shared irregularities may represent a single recent innovation.
(8) A further possible innovation of PNP is attested by the forms for the locative demonstrative bases meaning “here”, “there (near addressee)” and “there (distant)”. The PPN reconstructions are problematical because three putative subgroups—TO, SO, and EP—exhibit three different sets of forms for these meanings (though they agree in the distribution of the locatives, which are defined by their occurrence immediately following reflexes of the PPN locative-relational particles *ki “towards position” and *?i “relational position”). Once again, however, the external (non-Polynesian) evidence suggests that TO languages have retained the PPN forms, while NP languages have undergone irregular change.
The subgroup reconstructions are considered first. It will be noted that, in certain languages, the forms for the locative demonstratives are identical or nearly identical with the corresponding pronominal demonstratives (i.e., “here” and “this” are marked by one and the same form, as are “that” and “there”, etc. Cf. discussion of (2) above). In most - 271 NP languages, however, the formal similarity between locative and pronominal demonstratives is limited to the second morpheme (marking position) (relative to speaker or addressee) while the initial morphemes are quite different. It is suggested here that in PPN the same set of demonstrative forms occurred in both pronominal and locative functions, and that NP languages reflect an innovation of PNP in having distinct initial morphemes in pronominal and locative demonstratives.
PTO *heni ˜ hani “here”, *hena ˜ hana “there (near addressee)” and *hee “there (distant)”, are reconstructed from: TON, EUV heni, NIU hinai ˜ hinei “here”; TON, FUV hena, NIU hina ˜ hanaa “there (near addressee)”, RON, EUV hee “there (distant)”, NIU ee “this, these”. Note also PTO *koo from TON, NIU, EUV koo “there (yonder place)”, and NIU kunaa “there (yonder)”. The EUV and TON pronominal and locative demonstratives are morphemically identical: initial h- is a morpho-phonemic variable required following the locative particles ki and ?i. (The definite article exhibits the same variation—in both TON and NIU it is he after the locative particles, e elsewhere.)The NIU demonstratives differ from the corresponding locatives in the -i- present in the latter, a difference for which no simple explanation can be offered at present. NIU -nei ˜ nai, common to both the pronominal (e . . . nei “this, these”) and locative forms (hinei ˜ hinai “here”) can be derived from PTO (eni ˜ *ani by metathesis (cf. discussion of feature (2) above).
PEP *ko-nei “here”, *ko-naa “there (near addressee)”, *ko-raa “there (distant)” and *koo “there (yonder)” are reconstructed from: MAO, RAR konei, konaa, koraa, and koo and TAH ?onei, ?ona, ?ora, and ?oo,, respectively, and from cognates in EFU, NAN, VAI, TIK, and WFU cited below. Other EP languages for which we have information show inconsistencies. Members of Green's putative Marquestic subgroup (MQA, HAW, and MVA) reflect PEP *koo “there (yonder)”, but show differences from MAO, RAR, and TAH in the other locatives: MQA has nei “here”, naa ˜ ke?ina “there (near addressee)”, and ?aa ˜ ?ei?a ˜ ke?ika “there (distant)”; HAW has ne?i ˜ ?ane?i “here”, la ˜ ala ˜ laila “there (distant)” (all following the locatives particles i and ma).
EAS has nei “here”, ena “there (near addressee)”, and era “there (distant)”, forms which also occur in the pronominal demonstratives te . . . nei, te . . . ena, te . . . era, etc. (cf. discussion of feature (2) above). TUA has teia “here”, teenaa “there (near addressee)”, and teeraa “there (distant)”, but the distribution of these forms is unknown, though they are formally identical with the singular pronominal demonstratives.
Certain SO languages reflect *konei or *kunei, *konaa and *kolaa, in agreement with MAO, RAR, and TAH, a fact which together with other evidence noted below requires that these forms be reconstructed as PNP. These SO languages (all suspected, but not established, as belonging to a subgroup of SO) show the following forms: EFU, TIK, WFU kunei, NAN, VAI konei “here”, EFU, TIK, WFU kona, NAN, VAI konaa “there (near addressee)”, TIK, WFU kora, EFU kola, NAN, VAI kolaa “there (distant)”. The variation between kunei and konei is unexplained.- 272
Most other SO languages, however, reflect *kinei, *kinaa, and *kilaa: SAM ?inei, ?inaa, ?ilaa; PUK, TOK, SIK, NUK kinei, kilaa; REN kinei, kinaa, kigaa; LUA iŋei, iŋaa, ilaa. KAP has kinei “here” but kono and koro “there”.
Four Outliers appear to have developed new forms: PIL mua-nei, mua-na, and mua-la; FIL kenei ˜ kenei (others unknown); WUV di-nei, di-na,—; MAE nei, na, laa.
It can be seen that while they are not demonstrably cognate in their initial morpheme, the two main sets of locative demonstratives found in NP languages are at least structurally similar in that they consist of two morphemes (all reflect initial *ko- *ku-, or *ki- “location marker” plus final *-nei “near speaker”, *-naa “near addressee”, or *-laa “distant”). This bimorphemic structure is also found in the PIL, FIL, and WUV, though the initial morpheme differs formally in these languages. Only MAE, MQA, EAS and possibly the HAW locative demonstratives are monomorphemic, lacking an initial location-marker. All this strongly suggests that the PNP locative demonstratives were bimorphemic, consisting of a location-marker plus an affix marking position-relative to speaker and/or addressee. Moreover, because of their wide distribution in both SO and EP subgroups, we may posit *ko-nei, *ko-naa, and *ko-laa as the actual PNP forms. 31 The other widespread forms (reflexes of *ki-nei, *ki-naa, *ki-laa, and *nei, *naa, *laa) may be regarded as representing innovations of certain daughter languages of PSO or PEP. In the case of *ki-nei, etc., only the shape of the initial morpheme is changed, while the structure remains the same. In the case of *nei, etc., loss of the location-marker is assumed.
Comparison of PNP *konei, *kolaa, and PTO *heni ˜ *hani, *hena ˜ hana, *hee, shows correspondence of function but not form. Which (if either) series of forms regularly reflects PPN.
No definite answer can be given to this question on the basis of the internal PN evidence alone, but comparison with other, closely related Oceanic languages suggests that TO languages have retained the PPN forms (possibly with insignificant changes), while NP languages reflect a significant innovation of PNP, namely development of a special general locative formative *ko- prefixed to the markers of position relative to speaker and/or addressee. 32
Several other Eastern Oceanic languages resemble TON, NIU, and EUV in their treatment of pronominal and locative demonstratives. In these languages, as in TON, NIU, and EUV, the semantic distinction between “demonstrative pronoun” and “demonstrative locative” is marked syntactically rather than morphologically; i.e., the semantic distinction is marked not in the demonstrative forms themselves but by other features of the utterance: a positional or directional particle plus a demonstrative - 273 is interpreted as a locative, e.g., TON mei heni “from here, from this place”, ki heni “to here, to this place”; but following a nominal base or specifying particle, a demonstrative is given a pronominal interpretation as TON ko eni “this (one), these (ones)”, ko e falee eni “this is the house”.
From data of limited quality, it appears that the Southeast Solomon Is. languages Nggela, Bunotu, Kwara'ae, Lau, Longgu, etc., cited below have demonstrative constructions which correspond both in form and function to those of TON, NIU, and EUV. In the following examples from these languages the forms i, kei, tei are position-markers (“at”), while . . . marks position of nominal base or personal article:
Nggela - i ani ˜ i nei “here”, i ane “there”, . . . eni ˜ nei ˜ ini “this, these”. Kwara'ae i ne?e “here”, i meana “there”. . . . nini ˜ ne?e “here”, . . . nea ˜ neana “that, those”; Longgu i nei “here, now”, . . .ne ˜ nei ˜ nene “this”,. . .ine ˜ nina “that”; Marau Sound tei eni ˜ kei eni “here, at this place”, tei na “there, at that place”, . . . nei “this, these”, . . . na “that, those”; Bugotu i ani “here”, . . . eni ˜ ani ˜ ŋeni “this, these”; Fagani i anai “here”, i ai “there”, ta ini “now”, . . . ŋare “this” . . . ŋaea “that”; Sa'a -ni “emphatic demonstrative suffix”, -ne “general demonstrative suffix, this, that”, . . . ena “that”; Ulawa . . . ini ˜ ine “that”; Gao (?i) ani “here”, . . . ani “this”.
Cognates in a number of other widely dispersed languages have been noted: Malagasy ini “this, that”, Motu, Tagalog, Malay ini “this”; Cakaudrove (Vanua Levu, Fiji) ena “that, there (near addressee)”; Yasawa Is. (Fiji) ini-ake “that, there (near addressee)”. Kwamera te-ini ˜ ine ˜ nei “this”; Mota i ake “here”, i ane “there”, . . . ike “this”, . . . ine “that”.
From the foregoing evidence PEO *i eni ˜ ani ˜ ini “here”, i ena ˜ ina ˜ ana “there”, . . . eni ˜ ani ˜ ini “this, these”, and . . . ena ˜ ina ˜ ana “that, those” can be reconstructed. The TON and EUV pronominal and locative demonstratives can be derived regularly from these, with the exception of initial h- in the TON, NIU locative demonstratives: NIU. . .nai ˜ nei “this, these”, can be derived from *ani ˜ eni by metathesis. The initial h- of TON, EUV, and NIU locative demonstratives has no correspondences in other languages and may reflect an innovation of PTO.
It follows from this that the corresponding PPN demonstratives (i.e., those indicating position near speaker and addressee, respectively) must be reconstructed as identical with those of PTO, except for initial *h- in the PTO locative set. That is, PPN *i (h)eni “here”, *i (h)ena ˜ (h)ana “there (near addressee)”,* . . . eni ˜ ani “this, these” and* . . . ena ˜ ana “that, those (near addressee)”, are reconstructed. 33 The development of PNP *konei, *konaa, and *kolaa may be explained as resulting from extension of the domain of PPN *koo “position yonder, over there (indicated)” to include occurrence before *-nei, *-naa, and *-laa. In this - 274 position *ko- simply marked the construction as locative (while occurrence of *te in this position marked the construction as pronominal (cf. preceding discussion of demonstratives in NP languages)).
Seven other features, shared by all or may NP languages apart from TON, NIU, and non-PN languages have been noted. 34 For various reasons, it cannot be established whether these represent innovations or retentions from PPN. Certain of these features are found in Outliers, notably the following:
(9) PNP *fea “time-space interrogative, when? where?” is reflected by: SIK hea, LUA, KAP, NUK hee, PIL, TIK, TAK, WUV fea, FIL fea ˜ fia, WFU wafe, MAE fee. PTO *fee, reconstructed from TON, NIU fee, could reflect either PPN *fee or *fea. Since assimilation of PPN final *-a to -e following a front vowel occurs semi-regularly in TON and NIU, and in several words in KAP and NUK, it seems probable that all the above forms derive from PPN *fea.
(10) PNP *fua “just, only, merely, at random, without purpose”, a postposed marker, is reflected by several SO languages: NAN, VAI, EFU, SAM fua, TIK fua ˜ fuare, PUK wua. The only EP language with a probable cognate is RAR, which has a postposed marker ?ua “merely, only, just, without purpose”, but this is sufficient to require reconstruction of *fua as PNP. (The functional equivalent in TON and NIU is noa, which also occurs in many NP languages, and must be regarded as PPN.)
(11) PNP *lava “intensive, particular, very”, a postposed marker, is reflected by: SAM, EFU, TOK, PUK lava, MAO rawa, RAR rava. The only known Outlier cognates are WUV lava and WFU rava “self, own”.
(12) PNP *laa “distant from speaker and addressee”, a postposed marker substituting for *nei “near speaker” and *naa “near addressee” is reflected by all known NP languages, 35 e.g., SIK, LUA, NUK, WUV laa, KAP, TIK, MAE, FIL raa.
(13) PNP te?e-ai “not, none, to be lacking, non-existent” a bimorphemic negative reflected by MAO tee-ai, SAM, EFU lee-ai, KAP teeai, TOK heeai, LUA seai, PIL, TIK, SIK si-ai. (Cf. also (1) in Sec. 3).
3. The Samoic-Outlier Subgroup
Within NP, two major subgroups, EP and SO, are posited (Members were listed in sec. 2). The EP languages have long been recognised as forming a well-marked subgroup, sharing as they do a large number of lexical and morphological features, and certain phonological features, exclusively of the rest of Polynesia. 36 There is little evidence for including any Outlier in EP: the only innovation known to be shared by an Outlier with EP, exclusively of other Triangle languages, is the merger of PPN *s and *f, which occurs in KAP and LUA. In the absence of other - 275 evidence of subrelationship between these two languages and EP, this can be regarded as an instance of convergence.
There is, however, fairly good evidence for including all known Outliers in the Samoic subgroup posited by Elbert 37 and Pawley. This evidence consists mainly of uniquely shared morphophonemic and morphological features. (There are no known phonological innovations characterising Samoic and/or Outlier languages apart from EP, and while detailed lexical comparisons are in progress the results are not yet available. 38 The lexico-statistical figures obtained by Bayard 39 are suggestive, but no confident interpretation can be given them without other knowledge.)
Some 12 morphological features shared by members of the putative Samoic subgroup exclusively of EP, TO, and non-PN languages have been noted previously. 40 It was argued that two of these features probably represent shared retentions from PNP, while the rest may represent shared innovations of the subgroup ancestor (Proto-Samoic). The present decision to include the Outliers in a SO subgroup permits some revision and expansion of this list (which now refers to SO rather than to Samoic). All 16 features in this revised list are present in many SO languages, and in most cases their wide distribution necessitates the conclusion that they were present in the subgroup ancestor, PSO. These features are as follows: (1) PSO *se?e “not”, *se?eki “not yet”, replace PPN *te?e and *te?eki, respectively, as prenuclear negative particles, with irregular substitution of *s- for *t-. *se?e “not” is reflected by SIK, LUA, TIK, MAE, VAI, see EUV, REN he?e, PUK ctee NAN, TOK hee, SAM (archaic) see “not”, and *se?eki “not yet” by SIK VAI, TIK, seki, NAN, TOK heki, REN, EUV he?eki, LUA sei “not yet”; SAM se?i “deferential tense-aspect, until, etc.” (Cf. also fn. 40)
The PPN negatives *te?e “not” and te?eki “not yet” differ in having initial *t- rather than *s-. The PPN forms are reconstructed from the SO negatives cited above plus the following: Fijian (Bauan) dee “negative imperative”, NUK, KAP, MVA, MAO (archaic), MVA, tee “not”, NUK, KAP tiki, TON te?eki “not yet'.
The only putative SO members which show initial *t- rather than *s- are NUK and KAP. SAM, however, has a double set of reflexes: see and se?i (noted above), and lee “not” and le?i “not yet”. EFU negatives lee and leiki correspond to the second set of SAM forms in exhibiting initial l-. In both SAM and EFU this l- probably replaces an earlier *t-; the same change has apparently occurred in the singular definite article, - 276 where PPN *te is replaced by SAM, EFU le. It seems likely that the EFU article and negative particles have been borrowed from SAM, while the SAM forms lee and le?i may themselves represent an earlier borrowing from TON 41 The NUK and KAP forms are unexplained by the present subgrouping hypothesis.
(2) PSO *ŋaa- direction prefix to certain locative bases such as *tai “sea” and *?uta “land”, as *ŋa?auta “landwards”. Reflected by TOK, SAM, VAI, NAN, KAP, NUK, PIL ŋaa-, EFU, TIK ŋa-, FIL ŋali-.
(3) *soko- “only, restricted to, exclusively”, prefix to nuclear person-markers and/or numerals: SIK, TOK, NAN hoko-, LUA so- ˜ so?o, TIK soko-; PIL oko-; NAN hoko- “exactly”; cf. also REN hoko, NUK sokosoko “alone”, SAM so?o- “any” (prefixed to indefinite article plus numerals).
(4) *na “definite article plural” (rather than PNP *ŋa, which occurs as a plural article or as a post-article plural-marker in certain EP and SO languages) is reflected by REN, PUK, KAP, PIL na, TOK naa “def. article pl.”. Cf. also SAM naai, EUV, VAI nai, “diminutive article plural”. On the other hand, TIK, NAN, NUK, WFU, FIL, and MAE reflect *ŋa. Cf. also (16) below.
(5) PSO *oulua “2nd pers. dual, preposed possessor” and *outou “2nd pers. pl., preposed possessor” replace PPN *m(o)urua and *m(o)utolu, respectively. The evidence for these PSO reconstructions was given in sec. 2, in the discussion of PNP morphological innovation (1). It was claimed there that PNP *oulua and *outou arose from loss of the initial *m- in the PPN forms, a loss which also occurred in the somewhat different nuclear person-marker forms.
However, it is just possible that *oulua and *outou represent an innovation of PSO rather than PNP. This possibility is suggested because the corresponding preposed possessive pronouns in EP languages reflect *ko(u)lua and *koutou, with initial *k-. The latter forms are reflected by - 277 all NP languages (including SO) in the nucleus slot, where they mark subject or object, as well as possessor (see sec. 2), and it is possible that in reflecting *koulua and *koutou in both the possessive and nuclear forms, EP languages simply follow PNP. The weight of the evidence, however, is against this view. It seems more likely that PEP speakers generalised the nuclear forms to the preposed possessive position, thus eliminating an inconsistency in the total system, while SO speakers retained the PNP distinction between preposed and nuclear forms.
Nevertheless, it is important to realise that whichever solution is adopted, either a NP or a SO subgroup must be recognised. If *oulua and *outou, without initial *k- are regarded as PSO, but not PNP, this is strong evidence for a SO subgroup. But if, on the other hand, it is accepted that PNP developed *oulua and *outou, along with the nuclear pronouns *koulua and *koutou, after losing *m from the PPN cognates, then this is strong evidence for a NP subgroup.
(6) A similar case is provided by the indefinite articles in NP languages. PPN *ha “indefinite article” is replaced by PSO *se “indefinite article singular” and *ni “indefinite article plural”. The latter are reconstructed from: SAM, EFU, NUK, REN, TIK, VAI se, SIK, LUA, TOK, KAP, NAN, EUV he, TOK e “indefinite art. sing.” and SAM, REN, TIK, KAP, NUK, VAI, NAN, SIK, PUK, TOK, ni, LUA ŋi, EFU ni- ˜ ni?iki “indefinite art. pl.”. Note also WUV e isi ni “a limited hina “indef. art. Pl.”, ni ˜ ni?ihi “others”, TON ni?ihi “a limited number, several, some” (a numeral-class base as ha kato ?e ni?ihi several baskets).
EP languages, like TON and NIU, do not distinguish singular and plural indefinite articles. EP languages reflect *se in both functions (though, unlike TON and NIU, they distinguish between singular and plural definite articles). Again the question arises: do EP languages retain the older usage, or have they lost an earlier *ni “indef. art. pl.” and come to use reflexes of *se to mark both singular and plural? Again, if we claim that EP languages retain the older usage, we must admit a SO subgroup. The contrary position is argued here, however, though not as strongly as in the case of (5) above. The cognacy of the TON base ni?ihi with the indef. art. pl. in SO languages (plus the presence of a cognate morpheme isi in certain SO languages—see (13) below, suggests that *ni or at least *ni?isi occurred in PNP, though not necessarily as an article.
The fact that the distinction between singular and plural number is such an important structural feature of EP as well as SO languages, is an a priori reason for assuming that the distinction would have been made in the indefinite articles in the language ancestral to them, namely PNP. The fact that SO languages do make this distinction, with a form which- 278
has a cognate in TON, makes it even more likely. Enough uncertainty remains, however, enough to consider the *se/*ni distinction a possible innovation of PSO.
(7) In many SO languages the dual and pluralpre posed possessive pronouns for the first and third person show a peculiarity not matched in EP or TO languages. They consist simply of an article plus a person-marker, as VAI te-maatou fale “our (pl. exc.) house”, ø-maatou fale “our houses”. In EP and TO the corresponding pronouns contain a possessive infix (-a- or -o-, indicating dominance or subordinacy towards thing possessed), as MAO t-oo-maatou fare “our (pl. exc.) house”, ø-oo-maatou fale “our houses”. The SO forms can be labelled “neutral possessive pronouns” because they do not make the characteristic PN distinction between a and o categories of possession. Such neutral possessives (in the 1st and 3rd person dual and plural pronouns) occur in SAM, VAI, NAN, TOK, REN, SIK, KAP, and NUK. In SAM they co-exist with forms which do make the usual distinction; in some of the other languages the distinction is made following the definite article plural but not after the singular article.
WUV, EFU, and PUK appear to retain the PPN and PNP a/o distinction in all forms. This situation can be explained either by assuming that these three languages fall outside the subgroup containing those SO languages which have “neutral possessives”, or by assuming that they once had neutral possessives as stylistic variants of a/o possessives (as SAM but that the neutral possessives eventually lost out in competition with the alternant constructions).
MAE and LUA have undergone somewhat different changes in the non-singular possessive pronouns, which do not affect the issue here.
(8) *kai marking sequential or consequential relationship between actions and variously translated as “then, and, next, but, and then” is reflected as a subordinate verb phrase initiator by KAP kei SIK, TIK, NUK, WUV, REN kai, MAE, WFU kaie, and possibly by SAM ?ae, REN, EFU, NAN kae ˜ ka. Cf. also Fijian (Bauan) ka “verbal clause conjunction”.
(9) *koi “present progressive” is reflected as a preposed tense-aspect marker in verbal phrases by PUK, SIK, TOK, TIK, NAN, VAI koi. MAO kei, TAH, RAR tei locative particles marking “present position”, TON kei “still” and MAO koi (archaic) “whilst” are possible cognates but do not appear to distribute as tense-aspect markers.
(10) Two other particles marking progressive aspect are foundin SO languages. *noko “past” or “progressive” is reflected as a preposed tense-aspect marker by KAP, NUK, WFU (Aniwa dial.), REN noko “past progressive”, and possibly by FIL, WFU no “progressive”, SIK kona “past progressive” and SAM ?ona “dependent verb phrase initiator”, “because”. On the other hand, several SO languages, (notably those which do not have a certain reflex of *noko) reflect *ku or *ko “progressive aspect”: SAM ?o ˜ ?olo?o, TOK, PUK, VAI, TOK ko, NAN kaa-ko, EFU ku. Cf. also TON oku “progressive”.- 279
It seems likely that *ko or *ku is the older form; *noko perhaps arose by combining the past marker *na with *ko (in the order na + ko in certain languages and ko + na in others).
(11) PPN, PNP *toko- “human number”, is prefixed to numerals and the base *fia “how many?” is reflected by most members of the putative SO group as toka-, with irregular change of the final vowel from -o to -a.
PPN, PNP *toko- is reconstructed from: TON, NIU, EUV, MAO, MVA, MQA, EAS, NUK, KAP, MAE, NAN toko-, HAW ko?o-, TAH to?o- “human number prefix”. The following languages reflect *toka-: SAM to?a-, REN, WUV, EFU, VAI, PUK, SIK, TIK, toka-. 43 LUA has ko?o- before lua “two” and tolu “three”, ko?a- elsewhere. It would be possible to reconstruct *toka- as the PSO form, if it were not for the fact that MAE, KAP, NUK, and NAN, all putative members of SO, have toko- rather than toka-.
The distribution of these two variant forms can be explained in several different ways, (a) It can be assumed that MAE, NUK, KAP, and NAN fall outside the subgroup consisting of those languages which reflect *toka-. (b) These four languages once had the form toka-, but it became toko- through assimilation of the unstressed final -a to the preceding stressed -o-. (c) The subgroup ancestor contained both toko- and toka- as allomorphs, and most daughter languages selected one allomorph over the other. LUA alone retained both, (d) Borrowing has widened the distribution of one or the other variant, thus obscuring subgroup boundaries.
This is a good example of the kind of problem often faced in sub-grouping and reconstruction. Many alternative explanations of the data are possible in this case, and a choice can be made only on a probabilistic basis, by considering other evidence. In the present case, there is other evidence that MAE, NUK, KAP, and VAI belong in a subgroup with the nine languages reflecting *toka-, and explanation (a) runs counter to this evidence. Explanations (b) and (c) are both consistent with the SO hypothesis. Explanation (d) is consistent with any hypothesis and thus is not illuminating (though it may be true).
(12) PSO *tou “1st person inc. plural” is reconstructed as a stylistic variant of *taatou in the preposed slot, as *te tou fale “our house” varying with *te taatou fale and *too taatou fale (cf. (7) above). *tou is reflected by WFU tou, MAE tu “1st person inc. pl., preposed possessor and preposed subject”, EUV tou, WUV gi-tou “1st person inc. pl., preposed subject”, TIK too, VAI tou “1st person inc. preposed possessor”, SAM, SIK, LUA, KAP, NAN, NUK reflect *taatou only. (However, SAM and NAN have a form tou “2nd person plural, preposed subject or possessor” which would be homophonous, and thus in competition with such a form for - 280 1st person inc. pl.). Evidence from other putative members of SO is not available.
(13) Most SO languages reflect a morpheme *?isi, either as a verbal base meaning “to be, to have”, or as a particle following the nominal articles, and meaning “some other, additional”. It is possible that these two apparently different markers derive from a single morpheme in PSO: no language seems to have the form in both meanings and both distributions. SAM, MAE isi, TOK ihi, EUV ?ihi, REN ?isi, EFU iki occur after nominal articles as SAM se isi tamaaloa “another man”, ni isi tamaaloa “some other men”. SIK, LUA, VAI isi occur following verbal particles as e isi “there are (some)”, marking the opposite of e siai or e seeai “there are none”, “no”, “none” (cf. (14) below).
TON has a monomorphemic numeral (verbal) base ni?ihi meaning “(to be) several, a limited number of”, as ha kato e ni ?ihi “several baskets”, ?i he taimi e ni?ihi “sometimes”. This suggests that in PSO *ni?isi functioned as a base meaning “(to be) a number of, (to be) some”, and was inherited by one group of daughter languages as a base ?isi meaning “(to be) some”. A second group associated ni in ni?isi with the indefinite article plural ni (see (6) above), and treated ?isi as a separate morpheme meaning “an additional number, other”.
(14) *a “generic or plural article” is reflected as a preposed marker substituting for other nominal articles by VAI, EFU, TIK, MAE, KAP, NUK, PIL, WFU a “plural article”. The exact meaning of a in these languages is not clear; in PIL, MAE, WFU, FIL, a appears to mark generic category as well as plurality, and in most of these languages a coexists with both a plural definite and a plural indefinite article. Probable cognates occur in SAM, EUV aa “nominal specifying particle”, TON, NIU ha “indefinite article”.
(15) A quantitative nominal article *sina or *suna is reflected by SAM sina “some, quantity of mass noun” (as sina suka “some sugar”, EUV hina “indefinite article plural”, KAP huna, NUK, PUK hanu “some, a quantity of”.)
(16) A preposed nominal particle *nai or *naai is reflected by SAM naai, EUV nai “diminutive or affective article plural” (as SAM naai-?u uso “my dear brothers”, EUV nai kaiŋa “dear parents”), VAI nai “a few, diminutive particle” (as nai mu pati fua “just a few words”).
Probable cognates include TON ŋaahi, a particle which occurs following articles or possessive pronouns and marks plurality of nouns denoting people, as hono ŋaahi ?ofefine “his daughters”, and the “tribal” prefix of the EP languages RAR, TUA, and MAO. In MAO ŋaai- is prefixed to names of tribal groups, and also to certain pronouns and nouns to indicate a group of people, as ŋaai-maaua “I and my people”.
From this evidence PPN *ŋaahi can be reconstructed as a particle occurring preposed to nouns denoting people, and marking plurality.- 281
*ŋaahi probably occurred either following the article (as in TON) or as an article.
4. The Internal Relationships of Samoic-Outlier Languages
4.1. In Sec. 3 evidence was given in support of the claim that the NP group divides into two coordinate groups, SO and EP. The SO group includes all (linguistically) known Outliers, together with SAM, EFU, VAI, NAN, TOK, PUK and probably EUV. It remains to attempt to determine the subrelationships of the members of this putative SO group.
This is a task of considerable difficulty, not lessened by the limited quality of descriptions available for SO languages. Where relationships between languages are fairly distant it may be possible to determine subgroups on the basis of quite simple methods of comparison, using fairly gross descriptions. Where closely related languages are concerned, however, some subtlety is required. In the case of the SO group, all members are very similar; a detailed comparison of all aspects of these languages, particularly of grammatical structure, and of large lexical samples, is necessary to yield enough evidence of innovations to determine subrelationships.
A good illustration of the problems faced in such a study is provided by Green's recent account of EP subgroups. EP languages all show much the same degree of close similarity to each other as do members of the SO group, though in both cases there is reason to believe that some subrelationships can be established. This optimism is based partly on the unusual settlement history of Polynesia; in many cases islands must have been settled by small groups of accidental migrants whose descendants remained cut off from significant contact with other Polynesian-speaking communities. In Polynesia we scarcely find the development of dialect chains, with diffusion obscuring earlier differences and inhibiting further differentiation, which is so familiar on larger land-masses. This is particularly true of Eastern Polynesia, where most of the major island groups are separated by many hundreds, even thousands, of miles of ocean.
Green did find evidence of borrowing between TAH and MQA, and possibly TAH and HAW, and there is other evidence of borrowing between TAH and dialects of the TUA archipelago. In his opinion, however, several subgroups are still distinguishable. He posits an initial split of PEP into Proto-Central Polynesian and Proto- (or Pre-) EAS, the latter being ancestral only to EAS. Proto-Central split into Proto-Marquesic and Proto-Tahitic, the former being ancestral to HAW, S.E. MQA, N.W. MQA and MVA, the latter to TAH, TUA, RAR, and MAO. Within the Marquesic subgroup it is claimed that HAW and S.E. MQA form a subgroup apart from N.W. MQA and MVA. 45
Green's evidence for these subgroups is fairly slim. His Central subgroup “is established on the basis of one phonological and two lexical innovations, his Marquesic on the basis of 6 lexical innovations, his - 282 Tahitic on the basis of 3 lexical innovations.” 46 The limited quantity and quality of the evidence results from the fact that Green restricts his comparison to phonology and a restricted area of vocabulary; linguists cannot hope for really substantial proof of PN subrelationships until they attempt exhaustive comparisons. Such detailed study must await the appearance of good grammars (covering both phrase and sentence structure) and dictionaries for the many PN languages which still lack them.
In subgrouping SO languages the same problems as with EP are faced, plus the probability that more borrowing has occurred in members of SO. In general, SO communities have had far more extensive contact with each other, and with TO languages, than have the more widespread EP communities. TON, NIU, PUK, TOK, EFU, and EUV all lie with 400 miles of SAM. The central Outliers, TIK, SIK, LUA, TAK, NRA, NMU, PIL, and TAU form a second fairly closely-spaced chain, though they lie far to the west of the Triangle languages. The ECE chain, spread over a 600 mile span to the north of Rotuma, lies in a middle position between Triangle Polynesia and the central and northern Outliers. Four Outliers (MAE, WFU, WUV, and FIL) are found in the New Hebrides, where they are close neighbours of a large number of non-PN languages. Bayard 47 has claimed that almost all Outliers have had secondary settlement from or significant contact with other Outliers and/or Triangle communities (as well as with non-PN communities). Moreover, the linguistic histories of the nearest Triangle communities—East Futunan, Ellice, and East Uvean—also appear to have been complicated by secondary settlement and borrowing from other Triangle sources. 48
In view of the extent of borrowing among both Outliers and Triangle members of SO, we may ask whether it is possible to recover any of the subrelationships that must have once existed in the SO group. The answer seems to be that, borrowing notwithstanding, at least three SO subgroups are fairly well marked and several more are suspected on limited evidence. As noted earlier, one of the probable subgroups consists of the two northwestern Outliers KAP and NUK; a second consists of the central Outliers TAK, LUA, and SIK; a third consists of WFU and FIL, two of the New Hebrides Outliers.
Furthermore, there is some evidence that the members of the putative NUK-KAP and TAK-LUA-SIK groups belong in a wider group with the Ellice Is. dialects; this last suspected subgroup, however, requires further proof. Finally, the few features uniquely shared by EFU, TIK, WUV, FIL, and MAE indicate either subrelationship or borrowing.- 283
Something very like the last two suspected subgroups is implicit in Bayard's recent discussion of Outlier origins. He notes that the Ellice Is. and East Futuna are probable sources of most Outliers. 49 Only five Outlier cultures are not definitely considered to derive, directly or indirectly, from one or other of these two places: Rennell, Nukuoro, Pileni (of uncertain origins), Luangiua (said to be settled from Samoa or East Futuna) and Anuta (said to be settled from Tonga). Bayard remarks that derivation may be either directly by settlement from East Futuna or Ellice, or indirectly by settlement from another Outlier which was itself settled from these sources. He suggests that West Uvea was settled directly from East Futuna, while Tikopia was settled directly from either East Futuna or Ellice. Sikaiana was settled from Ellice, while Takuu-Nukuria was settled either directly from Ellice, while Takuu-Nukuria was settled either directly from Ellice or from Sikaiana. Nukumanu was settled from Takuu or Nukuria. Mae was probably settled from East Futuna, while Mele-Fila, West Futuna, and Aniwa “were almost certainly settled” 50 from Mae. In the case of FIL and WFU (including Aniwa) extensive borrowing from nearly non-PN languages has obliterated many clues to relationships with PN languages, though Bayard is fairly confident of their primary relationships. All the other Outliers, according to Bayard, have also undergone secondary settlement or had significant contacts with Triangle and/or other Outliers, resulting in considerable borrowing.
These conclusions seem to be based mainly on the lexico-statistical figures and kinship terminologies which he cites. However, in his discussion of “primary” and “secondary” sources, there is no indication of the particular linguistic items thought to be contributed by one or another source. No lexical or morphological innovations characterising particular suspected subgroups are cited.
The following subsections contain lists of those features believed to be uniquely shared by certain members of the SO group, together with brief discussion of the subgrouping or borrowing hypotheses which are suggested by these distributions.
4.2. Kapingamarangi and Nukuoro. KAP and NUK share a considerable number of morphophonemic innovations and lexical items apart from other known languages. With the exception of (1)-(6), the following are noted from basic vocabulary alone.
(1) Shortening of initial *-aa and loss of final *-a from PSO dual preposed possessive pronouns, as e.g.,
Loss of final *-a is not a regular change in KAP and NUK, but occurs in several other words, e.g., KAP iŋoo. NUK iŋo “name” (replacing PPN *hiŋoa), KAP, NUK hee “where?” (replacing PNP *fea).
(2) PPN *te?eki “not yet” (see (1) of sec. 4) are replaced by KAP, NUK tiki. with irregular change from *e . . . i to i . . . i. It is notable also that KAP and NUK are the only putative SO languages which retain PPN *t- in the negative particles *te?e “not” and *te?eki. VAI has siki as a stylistic variant of seki. Cf. f.n. 40 and (1) in sec. 3.
(3) KAP hoŋo. NUK hoŋa “on top, above”.
(4) NUK polo. KAP poroo “because, according to, that”, a phrase initial conjunction.
(5) KAP, NUK t-o-no “3rd pers. singular possessive, o- category”, replacing PSO *t-o-na. with assimilation of final *-a to -o. This form occurs only in some positions in NUK, which elsewhere has t-o-na.
(6) PSO *ŋaa- “directional prefix” (see (2) in sec. 3) is retained but with specialised meaning and changed distribution. In KAP and NUK *ŋaa- occurs before -iho and -ake. as: KAP ŋaake “up lagoon”, NUK ŋaake “in a clockwise direction around the circular reef”. KAP ŋaaiho ˜ ŋeiha “down lagoon”, NUK ŋaaiho “in a counter-clockwise direction around the circular reef”.
(7) KAP, NUK ka/mai “to give” (replacing PSO *kau/mai. as TIK, TAK kaumai. SAM, LUA ?aumai “to give”, with loss of *-u).
(8) KAP ŋoko. NUK ŋako “egg” (of chicken).
(9) KAP kore. NUK kole “seed”.
(10) KAP raani. NUK laani “day” (replacing PNP *laŋi with irregular change of *ŋ to n, and lengthening of *a to aa.
(11) KAP tuŋu. NUK tuŋa “to smell” (replacing PSO *soŋi. Cf. SIK, KAP sunu. LUA suŋu “to smell”.
(12) KAP taahiri. NUK tahili “to sing”.
(13) KAP roko. NUK lakolako “many”.
(14) KAP huiahi. NUK useahi “smoke” (replacing PSO *auafi or asuafi).
(15) KAP tehee, NUK tehe “how?” (formally corresponding to PSO *teehea “which?”, but semantically replacing PPN *peefea “how”).
(16) KAP taamaha. NUK taemaha “heavy” (replacing PSO *taumafa).
(17) KAP ŋahuru. NUK ŋaŋaulu “hair of head” (replacing PSO *lau ?ulu).
(18) KAP heehee. NUK sese “to walk”.
(19) KAP, NUK pakkau “wing” (suggesting an earlier *pakakau. replacing PSO *kapakau with metathesis of first two syllables).
(20) KAP roti. NUK lalati “to scratch”. Cf. TON lote “stir vigorously”, SAM lote “fidget”.
(21) KAP heepaki. NUK hepaki “to fight”. Cf. PPN *fepaki “to collide, clap together”.
These shared features require either the assumption of genetic sub-relationship, or of very extensive borrowing.- 285
4.3. Sikaiana, Takuu, and Luangiua. SIK, TAK, and LUA appear to form a well-marked subgroup. Data for NRA and NMU are not available, but other observers have noted that these are very similar to TAK, possibly dialects of the same language (see f.n. 4). As well as showing general structural and lexical similarity, SIK, TAK, and LUA share one phonological innovation and a very considerable number of morphophonemic and lexical innovations in basic vocabulary alone exclusively of other SO languages, including the following:
TAK and LUA also share certain features exclusively of all other known PN languages, including SIK. Since in many of these cases SIK has retained the PSO form, the TAK and LUA forms must be assumed to reflect either a shared history apart from SIK, or recent extensive borrowing between TAK and LUA. The limited evidence available for TAK precludes the drawing of any definite conclusions. The following uniquely shared features are from basic vocabulary alone.
For most of the following items, shared by SIK and LUA, comparable data for TAK are not available.
4.4. The Outliers and Ellice. The central and northern Outlier groups SIK-LUA-TAK and KAP-NUK uniquely share certain features with each other, and in some cases, with the Ellice dialects of VAI and NAN. However, the quantity and quality of probable shared innovations is considerably below that needed to establish a subgroup; borrowing must be considered as a possible explanation until further evidence comes to light. The following features occur in at least two of these three putative groups. (A few of the items cited also occur in PIL and/or TIK; these two languages are known to have had some contact with other central Outliers, but appear to group with EFU and the New Hebrides Outliers (see 4.5, 4.6), and may have borrowed these words from SIK, LUA, or TAK.
(1) The most important single innovation shared by SIK, LUA, TAK, KAP, NUK, VAI, and NAN is phonemic double (long) consonants. In most cases the double consonants (C1C1) reflect an earlier C1VC1, particularly when the medial vowel is unstressed or identical with the next vowel, e.g., PPN *lelei “good”, *sosolo “to wipe (plural subject)”, *totolo “to creep”, become VAI llei, ssolo, ttolo, respectively. LUA has very few double consonants, but appears to have taken the process a stage further, in that not only is the medial vowel lost but the double consonants are reduced to a single consonant, e.g., PPN *fafine “woman” becomes TAK ffine, LUA hiŋe, PPN *tinana “mother” becomes TAK, SIK tinna, LUA kiŋaa, PPN *mamafa “heavy” becomes SIK mmaha, - 287 LUA maha; TAK fakkatoa “all” (representing an earlier *fakakatoa) is matched by LUA ha?akoa.
The fact that these Outliers and all Ellice dialects with the apparent exception of Nitao 51 share this important innovation is the strongest single item of evidence for subgrouping them. However, the possibility of convergent development must be considered. The morphophonemic processes characteristic of other SO languages indicate that convergence in this case is by no means improbable. Most other SO languages so far observed show phonetic consonant clusters in fast speech representing just the same phonemic structures that reduced to phonemic double consonants in VAI, NAN and the central and northern Outliers. 52 Presumably this was a characteristic of PSO, and certain daughter languages may have independently extended the habit to all styles of speech.
4.5. The Outliers and East Futunan. The four New Hebrides - Loyalty Outliers (WFT, FIL, MAE, and WUV) share a few innovations with TIK and/or EFU exclusively of other SO languages (with the possible exception of PIL for which data are very scanty). The distributions of the following items are noteworthy:
It is also notable that WFU, MAE, FIL, and TIK all have a definite article plural with the shape ŋa, whereas REN, KAP, PUK, TOK with na “definite article plural”, SAM with naai, and VAI, EUV with nai diminutive article plural” all show a dental rather than a velar nasal in corresponding articles. WUV and EFU have no cognates. The *ŋa forms are probably a shared retention (see item (16) in sec. 3); if so, those languages which show a dental nasal share an innovation.
4.6. Futuna-Aniwa and Mele-Fila. WFU and FIL, besides sharing a few items cited in 4.5, have in common a considerable number of morphophonemic and grammatical innovations strongly suggesting subrelationship. Capell, who compiled the two brief grammars which are the only substantial published materials on these two languages, also commented - 289 on their similarity. 53 It seems likely that this subgroup will be confirmed when fuller lexical data appear for WFU and FIL. The following shared innovations have been noted to date:
Both have a singular definite article te ˜ ta ˜ ti contrasting with a dual article (FIL ru, WFU ru ˜ ruh), a trial article (FIL o, WFU taka ˜ tah), and a plural article (FIL a, WFU a ˜ a-ŋa ˜ ŋa). A few other PN languages show a three-way contrast between singular, dual and plural articles, with cognate forms (e.g., REN, NUK, MAE) but no other known PN language has a trial article.
(10) WFU, FIL avau “1st pers., nuclear” and WFU aha, FIL afa “what?” are matched in certain EP languages, but are unusual reflexes of PPN *au “1st pers. sing., nuclear” and *hafa “what?” among SO languages. No other SO language shows the intrusive -v- between the personal article a and the 1st person singular marker au. NUK (which has aha as an alternant of aa) is the only other non-EP language to retain the medial *-f- of PPN *hafa “what?” (note TON, NIU haa, EUV, SAM, EFU, TOK, PUK, REN, VAI, NAN, SIK, TAK, LUA aa).
(11) In the demonstrative constructions cited in (2) of sec. 2, WFU and FIL show an intrusive -ŋa- before the position marker, as WFU, FIL te . . . ŋa-na “that (near addressee)”, WFU ŋa . . . a-ŋa-nei, FIL a . . . ŋa-ne “these”.
(12) FIL, WFU fe- “desiderative prefix”, replacing PSO *fia, as FIL fe-unu, WFU fe-inu “to be thirsty”.
(13) WFU kamoa, FIL kaa(moa) “to hold”.
4.7. Rennell-Bellona. There is no substantial evidence at present connecting REN with any other SO language. However, there are certain items which REN shares with other languages which deserve comment.
(1) REN is the only Outlier language to retain PPN *?. Among Triangle languages *? is reflected regularly by TON ?, fairly regularly by EAS and EUV ?, and sometimes by EFU ?, and is lost in all other languages.
The Rennellese have traditions of a homeland called ?ubea. They also have traditions of contact with Pileni. The following items shared with WUV and PIL suggest some contact with these places. 55
(2) REN, WUV manaha “garden, place”.
(3) REN hale, WUV fale “kind, sort”.
(4) WUV efa, REN ?eha “big”. Note also PIL laueha, NAN lauefa, REN gau?eha “wide”.
(5) REN hetaiaki, WUV taiaki “near”. Cf. SAM taai, VAI taipili, PIL lavethaki, TIK rafitaki “near”.- 291
(6) REN ?oti, PIL osi, WFU oti “all”. These forms are cognate with PPN *?oti “finished” but semantically correspond to PPN *katoa “all”.
(7) REN gaoi, PIL lavoi, TIK laui, SIK, LUA laaoi “good”.
TEN also shares the following grammatical items exclusively with certain Outliers. The wide distribution of these items suggest that they are shared retentions from PSO rather than shared innovations.
(8) REN gu-, WFU, FIL ru-, NUK lu- “dual article”, occurring in the article slot in preposed possessive pronouns, as NUK lu-o-ku “my two, two of my”. A possible cognate is TON o-ŋo “dual marker”, which occurs postposed to articles or possessive pronouns, as e oŋo fale “the two houses”, and which may reflect an earlier (PPN) *ru-ŋa. (Cf. TON ua “two” (reflecting PPN *rua)) and TON -ŋo- “numeral stem formative” as in ho-ŋo-fulu “ten” reflecting PPN *ha-ŋa-fulu “ten”).
(9) REN, NUK, KAP, WFU (Aniwa dial.) noko “past” or “continuous”. (See (10) in sec. 3 for more detailed discussion).
(10) REN, NUK, KAP, WFU, FIL, PIL, share a possessive construction consisting of ni “possessive marker, belonging to” plus a/o- category possessive article plus noun or person-marker, as ni-o-ku “mine, belonging to me”. This ni is almost certainly cognate with n- “possessive marker” found in EP languages, as MAO, RAR, TUA, MVA n-oo-ku “mine, belonging to me”.
4.8. East Uvean. EUV has generally been grouped with TON with which it shares its highest percentage of basic vocabulary cognates, 56 and some characteristic grammatical features. However, in other respects it shows much more substantial resemblances to SO languages, and as can be seen from sec. 2 and 3, shares many features with NP and SO languages exclusively of Tongic. The presence of a number of doublets 57 and irregular reflexes of PPN *?, *h and *r 58 confirm that modern EUV is the result of mixing of two different speech traditions, one SO and one Tongic. In view of this, perhaps modern EUV cannot be meaningfully assigned to any genetic subgroup of PN. But it is reasonable to ask which speech tradition is dominant in modern EUV, and which can be considered the minor or borrowed component. Satisfactory treatment of this question would require a more detailed review of the evidence than is possible here. However, the following arguments are put forward briefly in support of the view that EUV is basically a SO language, overlain with recent borrowing from TON.
First, if EUV is a Tongic language it must have separated from TON only very recently, because those features it shares with TON include - 292 many which are not found in NIU and which appear to be recent innovations postdating the splitting-up of PTO, e.g., TON, EUV nima, NIU lima “hand”; TON, EUV naua, NIU laua “3rd pers. dual, nuclear”; TON, EUV toma?i, NIU matua taane “father”; TON, EUV kui, NIU tupuna “grandparent”.
If this were true, then it would require that the SO component be borrowed even more recently, presumably from some SO language in the western Polynesian area. But all the evidence is against this. The similarities of EUV are not to any particular SO language, but suggest more or less equal relationship to all other members of the SO group. That is, there is no large number of special resemblances to any particular SO language. Rather, there is just the distribution of shared features which would be expected of a SO language which had undergone independent development for many centuries. Further, EUV has some features which, though reflecting PNP or PSO features, can hardly have been borrowed recently from any nearby SO language (such as EFU, ECE, or SAM). For example, PPN, PNP *? is lost in ECE and SAM, and is not present in EFU cognates of the following EUV items: he?e “not”, he?eki “not yet”, ta?aku and to?oku “1st pers. sing, possessive” (and all other possessive pronouns); some or all of these forms are matched exactly in such distant NP languages as REN and EAS (but not in TON) and suggest that EUV *? is inherited from PNP, not borrowed from TON. EUV also reflects the PSO possessive pronouns *oulua “2nd person dual” and *outou “2nd pers. plural” as ulua and utou; the EFU and SAM equivalents are kulu and kotou ˜ koutou, and lua and tou, respectively, and could not have been the source of the EUV forms. Similarly, EUV has te “definite article singular” reflecting PNP *te, where EFU and SAM have le (cf. fn. 40).
All the evidence, then, is consistent with what one would expect if a SO speech tradition had undergone fairly extensive borrowing from TON sometime in the last few centuries. The system, despite the presence of a number of alien features matched exactly in modern TON, is still dominantly SO in type, but without particular resemblance to any other SO language.
4.9. Conclusion. No subgroups of SO containing more than three languages have been confirmed. The distribution of certain probable innovations noted in 4.4 and 4.5 is certainly suggestive of wider relationships, on the one hand, between the Ellice Is. dialects and several Outliers, and on the other between EFU and certain Outliers, but in neither case is the weight of evidence adequate to establish genetic subrelationship.
Difficulties arise in subgrouping SO languages because of the long history of contact between languages in the western Triangle area, and between neighbouring Outliers. In particular it seems likely that both EFU and the Ellice dialects have been extensively influenced by other Triangle languages, especially SAM (see fn. 20 and 40). If it is true that most Outliers derive from East Futuna and the Ellice Is., it may well be that these Outlier languages retain more of the subgroup ancestor or - 293 ancestors than either modern EFU or ECE do. For example, there are certain features shared by KAP and NUK (in the extreme north-west), REN (in the extreme south-west), and WFU and FIL (in the extreme south) that are found nowhere else (see 4.7). It is unlikely that borrowing could have occurred between these languages and it is also unlikely that these three groups of peripheral Outliers belong to the same subgroup of SO. This is so not only because of geographical barriers to voyaging between these areas but also because these languages share few of the special similarities expected of closely related languages. The simplest explanation is that the uniquely shared features are shared retentions from PSO or one of its immediate descendants. These features have since been lost or obscured in other SO languages. Particular importance in reconstructing PSO, therefore, attaches to the more isolated Outliers.
Little has been said of the possible relationships of SAM, PUK, TOK, and EUV within the SO group. In the case of EUV special difficulties arise because of borrowing, probably from TON (see 4.8), while the lexical data presently available for PUK and TOK (see fn. 1) are limited. Even so, it may be that SAM, EUV, PUK, and TOK each separated from other SO languages at some fairly early point. This seems likely not only on linguistic grounds but because these languages are all spoken within relatively short distances of each other in an area in which the main land masses—Fiji, Samoa, and Tonga—were probably settled before the breakup of PSO. The nearby smaller islands, such as East Uvea, East Futuna, and the Tokelau and Ellice groups, must have been settled fairly early by SO-speaking peoples; had they arrived only recently, such peoples would have found these islands already populated, as they did Rotuma and the Gilberts. 59 The geographic characteristics and early settlement of the Western Polynesian area, therefore, perhaps account for the fact that the SO group, although on present evidence a genuine subgroup, is certainly not as well marked as the EP group. The PEP-speaking community was clearly isolated for several centuries before it dispersed, and in that time developed a striking number of linguistic innovations. 60 At this time, after the breakup of PNP, the conditions for such isolated development must still have existed in some parts of - 294 Eastern Polynesia. It is unlikely that such conditions still existed in the Western Polynesian - Fiji area, where PSO was presumably spoken.
1 I am grateful to D. S. Walsh and Bruce Biggs for detailed comment on and criticisms of a draft of this paper, and to Anne Thorpe and Peter Ranby for checking the manuscript for errors and omissions. Unpublished field notes from research in progress as part of the Polynesian Culture History project of the Bernice P. Bishop Museum (financed by National Science Foundation grant No. GS/903, and by grant No. A/11 of the South Pacific Research Committee of the University of Auckland from funds made available by the Scientific Research Distribution Committee of the Golden Kiwi Lottery Funds) have generously been supplied by Peter Sharples (for Sikaiana), Rangi Moeka?a (Pukapukan), Anne Thorpe (Luangiua), Peter Ranby (Nanumea), K. J. Hollyman (for the West Uvean dialect spoken at St. Joseph Village (Heo), northern West Uvea), and Pat Hohepa (Niuean). Dr. Hohepa and Judy Huntsman have also kindly furnished unpublished notes on Tokelauan. Vera Carroll and Roger Green have contributed valuable ideas during discussion of some of the problems treated here.
2 This number is approximate for several reasons. First, little is known about the languages spoken on Anuta, Niuafo?ou, Taumako, Nukuria, Nukumanu, the Austral Is., and parts of the Tuamotu and Ellice groups. Second, it is often difficult to decide whether two closely related languages spoken on widely separated islands should be regarded as forming one language or two. If one takes less than 80 per cent agreement in basic vocabulary, or a considerable degree of mutual unintelligibility, to indicate language boundary, then there are at least 16 languages in Triangle Polynesia and at least 12 in the Outlier area (listed below). In addition to these it seems likely that Niuafo?ou, Anuta, and the Austral Is. will turn out to have distinct languages.
3 As well as the unpublished sources cited in fn. 1, I have made extensive use of text materials for several languages for which there are no or only sketchy published grammars, particularly REN, TIK, PIL and VAI. Sources are cited in the bibliography.
4 Bayard:15-16, found NMU and NRA to be dialects of the same language as TAK, according to lexicostatistical comparison. TAU appears to be closely related to PIL (Elbert 1965:439).
5 The few ANU items cited in Firth 1954 (the only source) include three words which strongly suggest close relationship to or contact with TON or EUV: tama. “father”, pae “mother”, kovi “bad”. Anuta traditions claim fairly recent settlement from Tonga, with later arrivals from Nanumea and, of course, from nearby Tikopia (Firth 1954:121-123).
6 Elbert 1953:169, Table 4; Pawley 1966a.
7 EUV has generally been regarded as most closely related to Tongan. It is argued here, however, that EUV is basically a Samoic-Outlier language which has undergone extensive borrowing from Tongan (see 4.8).
8 Elbert 1953:169, Table 4, Pawley 1966a:sec. 4.
9 NAN and VAI can probably be considered dialects of one language (ECE), but see discussion in 4.4 and fn. 20. PUK, despite some borrowing from Cook Is. languages of the Rarotongan type, is clearly a Samoic-Outlier language (see sec. 3).
10 Elbert 1953:169, Table 4; Emory 1963:89-90, Green 1966:34, Table 9. Elbert and Green also agree in grouping MVA and MQA together apart from a second group (Tahitic) which includes TAH, RAR, TUA, and MAO. However, Green differs from Elbert in including HAW in the same group (Marquesic) as MVA and MQA, rather than in Tahitic. Elbert and Emory appear to base their classifications of EP languages mainly on lexicostatistical evidence; Green also considers some phonological and lexical innovations (summarised in his Tables 2, 4, 6-8, and Appendix).
11 Elbert 1953:163-169. Elbert did not speak of a NP group but the grouping implied by his label P-Sa-O-E for the language said to be ancestral to SAM, ECE, TIK, KAP, and the EP languages is essentially equivalent to NP as defined here.
12 Pawley 1966a:sec. 2.1.6.
13 Bayard 1966.
14 Elbert 1965:434-441.
15 Capell 1962b:46. There are several earlier publications in which he has questioned the notion that all Outliers derive from Triangle Polynesia (Capell 1931, 1958, 1962a). A similar position is taken by Marshall 1956:66, 70.
16 Capell 1962b:46.
17 See Capell 1962a:391-392, 1962b:42-46, and Pawley 1966c for some discussion of this evidence.
18 The sole morphological feature which certain Outliers but no Triangle languages appear definitely to have retained from some earlier (PPN or pre-PPN) stage are two suffixed possessive pronouns attached to a very few kinship terms. All PN languages retain a reflex of PPN *-na which in PPN or pre-PPN marked 3rd pers. singular possessor following such bases as *tama- “father” and *tina- “mother”. In all Triangle languages the reflex of *-na is no longer meaningful and can be treated as part of the base. MAE, REN, PIL, FIL, TIK, and WFU also show a reflex of *-u “2nd pers. Sing. possessor”, while FIL and WFU also reflect *-ku “1st pers. Sing. possessor”. FIL and WFU also have suffixed dual and plural possessive pronouns, but these appear to be recent developments. These dual and plural pronouns are not added to the base, but to the base plus -n-(-n- plus a/o possessive in WFU), reflecting the *-na suffix, and are thus not structurally equivalent to the possessive pronoun suffixes of such related non-PN languages as Fijian. See item (5) in sec. 4.6 for further details.
19 This borrowing is indicated by irregular phoneme correspondences between cognates, and by the clusters of lexical innovations which are shared by languages in certain geographic areas which otherwise seem to belong to different subgroups. An example is the term kave “sibling of the same sex” which occurs only in TIK, ANU, FIL, WFU, SIK, LUA, AND TAK among known PN languages. It is not likely that all or even most of these languages belong to a single subgroup of SO, but they all lie in a central Outlier geographic area in which every Outlier is known to have some contacts with one or more other Outlier communities in the area.
20 This is suggested by the fact that certain ECE dialects share innovations with some Outliers while other ECE dialects retain the corresponding older (PPN or PSO) form. For instance, several ECE dialects share the feature of double consonants replacing earlier C1VC1 sequences with loss of medial vowel with the Outliers NUK, KAP, TAK, and SIK (see sec. 4.4). But at least one ECE dialect, Niutao (Kennedy 1946:6) retains the medial vowel in such cases. As these Outliers share several other features with ECE (see 4.4) it seems likely that they derived from one (or more) dialect areas in the ECE chain at a time when these dialects were already distinct from Niutao. A similar case involves the construction e isi “it exists, there is/are” which occurs in SIK, LUA, and VAI. The equivalent construction in PPN and PSO is *e iai which is retained by almost all other SO languages including the northern ECE dialect of NAN (which has e ai). VAI has both e isi and e iai as synonyms, which suggests dialect mixing.
21 Pawley 1966a.
22 By morphophonemic innovation I mean an irregular change in the shape of an item which is otherwise clearly cognate with the earlier (reconstructed) item, e.g., the 2nd pers. dual and plural markers discussed in feature (1) below. Generally, such an innovation provides better evidence of subrelationship than a uniquely shared word which has no cognates at all in languages outside of the suspected subgroup; in the latter case it can always be argued that the uniquely shared word is a shared retention, not an innovation.
23 In the orthography adopted here the velar nasal is written ŋ and glottal stop is written?. REN g represents a prenasalised voiced velar stop [ŋg]. NUK p, t, k represent voiceless stops which are written b, d, g in standard NUK orthography. In TON, NIU, EFU, WFU, PIL, and FIL t has affricate or fricative allophones before i which are written s or j in the normal orthographies. Loss of a reconstructed phoneme is shown by Ø.
24 This is also true of other closely related Oceanic languages, e.g., Bauan (Fijian) na . . . oqoo “this, these” as na kaa oqoo “this thing, these things”, where the demonstrative does not mark number.
25 Capell 1958:92, also lists enei, ena, and era as alternants of the plural demonstratives aŋanei, aŋana, and aŋara. No examples of the former are given, however, so that it is not clear what their distribution is.
26 Not cited in his grammar, but noted in texts in Capell 1962b.
27 See Pawley 1966a:sec. 3.2.7, and also in item (2), sec. 2 of this paper.
28 The terms “Fijian language” and “the Fijian dialects” are misleading. The “Fijian dialects” clearly include several different languages. Although continuing contact between neighbouring speech communities has probably limited diversification, the structural diversity of the Fijian languages and dialects is at least equal to that of the Polynesian group as a whole.
30 Such as Kwara'ae, Nggela, and Sa'a, for which fairly good descriptions are available.
31 Note also REN konei “like this”, konaa “like that”, kogaa “like that”, but kinei “here”, kinaa “there (by addressee)”, kinaa “there (distant)”.
32 *koo “yonder place” is, of course, PPN, but *ko- “locative demonstrative formative”, without other specific meaning, is found only in certain EP and SO languages.
33 Several south-east Solomon Is. languages have alternant forms in which the initial vowel varies according to the preceding vowel.
34 Pawley 1966a:sec. 3.1.
35 If PPN had a form *era “3rd pers. demonstrative”, to match *enei and *ena, its reflex in TON and NIU would be either ee or ea. In fact, TON and NIU do have a form ee with this meaning. However, TON, NIU ee “3rd pers. demonstrative” may be cognate with PNP *ee “antecedant pronominal marker”, as in SAM l-ee, MAO t-ee “he who, the one who . . .” and SAM Ø-ee “they who, the ones who . . .”.
36 Elbert 1953:163-169; Pawley 1966a:sec. 4.2.
37 Elbert 1953:169, Table 4, posits P-Sa as the language ancestral to Samoan, Ellice' and Tikopian.
38 Walsh and Biggs 1966, contains the first published results of comparative lexical research now in progress at the Universities of Auckland and Hawaii.
39 Bayard:52, Table 3, cites percentages of shared basic vocabulary for some Outliers with each other and with EFU, ECE (represented by VAI), SAM, EUV, TON, and HAW. Of the Outliers, TAK, LUA, SIK, TIK, MAE, and WUV show considerably higher percentages with EFU and/or ECE than with SAM, TON, HAW, or EUV. The figures for WFU, FIL, and NUK show no notably high percentages with any other SO languages; in the case of other Outliers the complete set of figures are not provided.
40 Pawley 1966a:sec. 4.3.
41 SAM has both lee and an archaic form see “not”, and both le?i “not yet” and a verbal particle se?i marking action which is unrealised but imminent or desired (often translated “just wait until, just let (me do . . .), until”). In view of the change, not paralleled in any other PN language, in the meaning of se?i, it seems unlikely that see and se?i are recent borrowings. It is possible, however, that lee and le?i have been borrowed. They correspond to EFU lee and leiki. Both the SAM and EFU initial l- in these particles appear to replace an earlier t-; that is, lee replaces earlier *te?e and le?i and leiki replace earlier *te?eki. Precisely the same change has occurred in the singular definite article in these two languages, where the free form alternant *te has become le, and the bound alternant *t- (which occurs in possessive pronouns and before -ee “antecedant pronominal marker”) has become l-. However, there is some evidence to suggest that EFU, like SAM, at some earlier stage has *se?e and *se?eki as negative particles, and that lee and leiki represent recent borrowings. This is suggested by the fact that those languages which seem to be most similar to EFU, namely TIK, WFU, WUV, and MAE (see sec. 4.5) all reflect *se?e and *se?eki rather than the PPN forms *te?e and *te?eki. The only language in the western Triangle area which has t- rather than s- or h- in the negative particles is TON, which has te?eki “not, not yet”. If SAM lee, le?i and EFU lee, leiki are not shared retentions, then TON seems to be the only likely source for these forms. Either SAM or EFU could have borrowed them from TON and transmitted them, with l- replacing t-, to the other.
42 As the preposed subject-marking pronouns have been lost in EP, KAP, and PUK, the nuclear forms *koulua and *koutou reflected by these languages would be in competition with only one other set of alternants, namely the (preposed possessive) reflexes of *oulua and *outou. The chances that one set of alternants will become dominant and that the other set will disappear are obviously higher in such cases than were there is also competition from a third set of alternants, as in those languages which also retain the preposed subject pronouns.
43 Although Durrad's wordlist contains TIK toko-, Firth in numerous texts gives only toka- and it is assumed here that this is the correct TIK form. Durrad's list contains several other errors in which MAO forms are wrongly attributed to TIK, and this may be the case here. Kennedy 1946, gives toko- as the VAI form, but I obtained toka- from an informant from Vaitupu.
44 Green 1966.
45 Green 1966:34, Table 9.
46 Biggs 1967.
47 Bayard n.d.:ch. 4-6.
48 See sec. 4.8 for brief discussion of borrowing in VAI, and f.n. 40 for a case of possible EFU borrowing from SAM or TON. Irregular phoneme correspondences in many other items in VAI, NAN, and EFU suggest further borrowing from nearby languages. In view of the history of contact between western Polynesian communities it seems likely that the more isolated Outlier languages, such as REN, WUV, KAP, and NUK, will prove better witnesses in reconstructing PNP and PSO, and in reconstructing the earlier languages spoken on East Futuna, East Uvea, and the Ellice Is., than will these modern Triangle languages themselves.
49 Bayard: 84 et seq.
51 Kennedy 1946:6. See also f.n. 20.
52 This is the case in SAM, LUA, SIK and in the Matahenua (Bellona) dialect of REN (Elbert 1965:436).
53 Capell 1942:153, and 1958:165.
54 Capell 1942:161.
55 Items (2), (3), and (4) were noted in Elbert 1965:435.
56 Elbert 1953:158 gives the following percentages of cognates in a 200-word list (using the multi-cognate method of scoring): EUV-TON 86, EUV-NIU 72, EUV-EFU 83, EUV-TIK 78, EUV-SAM 70, EUV-MAO 61, EUV-HAW 55, TON-NIU 64.
57 For instance, hifo and ifo “downwards”, hake and ake “upwards”, te and (in certain constructions) e “definite article”, lua (free form) and ua (in compounds such as ua-fulu “twenty two”)
58 EUV reflects PPN *? sometimes as ?, less often as Ø, PPN *h sometimes as Ø, less often as h, PPN *r usually as l, occasionally as Ø.
59 Biggs 1965 demonstrates extensive Rotuman borrowing from at least two PN languages, one of Tongic and one of Samoic-Outlier type. Gilbertese also shows signs of considerable borrowing from a PN language of Samoic-Outlier type. As in Rotuman, this borrowing is indicated by the presence of two sets of phoneme correspondences to PN languages; diagnostic members of one set never occur in the same word as diagnostic members of the other set. In one set, occurring in the majority of cognates, the corresponding phonemes are often phonetically quite dissimilar, as Gilbertese /r/, TON /h/ and SAM Ø in Gilbertese tari, TON tehi-na “younger sibling of same sex”, SAM tei “younger sibling”. In the other set, occurring in a minority of cognates, the corresponding phonemes are often phonetically suspiciously similar. Moreover, these cognates are often items that occur in Gilbertese and in PN or certain PN languages, e.g., Gilbertese /r/ and PIL, TIK, SIK, LUA /l/ in Gil. raoi, PIL lavoi, TIK laui, SIK, LUA laaoi “good”. If one excludes such suspiciously similar cognates Gilbertese and PN seem to be only distantly related.
60 A point well made by Elbert 1953:165-166.