Reconstructing the Case-Marking and Personal Pronoun Systems of Proto Austronesian

Streams Converging Into an Ocean, 521-563 2006-8-005-021-000082-1

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

Malcolm Ross The Australian National University

Since Blust presented his reconstruction of Proto Austronesian and Proto Malayo-Polynesian personal pronouns in 1977, more data relevant to their reconstruction have become available. This paper takes account of relevant publications since 1977 and sets out a fresh reconstruction of Proto Austronesian personal pronouns, with supporting data from Formosan languages and interpretive arguments. Since personal pronoun systems in Formosan languages often incorporate the case-markers more generally used in noun phrases, and it is impossible to interpret the histories of the pronouns without taking account of the case-markers, a reconstruction of Proto Austronesian case-markers is also presented here. The goal of Blust’s 1977 paper was to show that all Austronesian languages outside Taiwan are characterized by certain innovations in their personal pronoun system, thus providing evidence for the Malayo-Polynesian subgroup of Austronesian. Blust’s findings are confirmed and augmented in the present paper.

Key words: Proto-Austronesian, reconstruction, pronouns, case-markers

1. Introduction

Blust (1977) reconstructs the pronoun system of Proto Austronesian (PAn) as part of his account of the internal subgrouping of the Austronesian language family. He describes pronominal innovations that occurred in Proto Malayo-Polynesian (PMP), the ancestor of all non-Formosan Austronesian languages, and that are reﬂected in its many daughter-languages. In the years since 1977, more data relevant to the reconstruction of PAn pronouns have become available, and in Ross (2002a) I published a revised reconstruction as part of a sketch of PAn morphosyntax. This reconstruction builds on the insights in Blust (1977) and on the work on pronouns in Dahl (1973) and Harvey (1982), as well as on advances in our understanding of the structure of PAn clauses originating largely in Starosta et al. (1981). In addition to new descriptions of Formosan languages, I have been able to draw on two papers on the typology of Formosan pronoun systems, namely Li (1997a) and Huang et al. (1999).

Malcolm Ross

My 2002 reconstruction was published with almost no supporting data or arguments, and in the present paper I seek to fill that gap. I dedicate this paper to my friend Paul Jen-kuei Li, without whose tireless labour and leadership the data for reconstruction would not be available. Professor Li has himself written about case-markers and pronouns (Li 1978, 1995, 1997a), as well as providing data in his descriptive works (e.g. Li 1973, Li and Tsuchida 2001, 2006). Less obviously, but just as significantly, he has inspired and enabled many younger scholars to describe Formosan languages. In the course of preparing this paper, I have examined recent accounts of pronoun paradigms for each Formosan language, paying particular attention to the functions performed by each pronoun set. To this end, I have preferred to use the fullest grammatical descriptions that I could find. My data survey is limited to Formosan languages. The reconstruction of PMP that I have used here was done some twelve years ago. I paid careful attention to form, but perhaps not as much attention to function as in my Formosan survey. To update the Malayo-Polynesian survey (for which organised supporting data have not been published) would be a gargantuan task, as it would entail looking at all available descriptions of non-Formosan Philippine-type languages. Blust (1999b) has argued that ten primary subgroups of Austronesian can be identified: nine in Taiwan (the so-called Formosan languages) and one outside Taiwan (Malayo-Polynesian). This paper is based on a thorough survey of the first nine subgroups and a somewhat less reliable examination of the tenth. In an alternative subgrouping, Sagart (2004) proposes that Austronesian has (ignoring extinct languages) only three primary subgroups: the Formosan languages Pazeh and Saisiyat are primary subgroups, and all other Austronesian languages fall into a group which Sagart labels ‘Pituish’. PMP is a fourth-order subgroup of Pituish and is even less significant for PAn reconstruction than under Blust’s proposal.1 Formosan languages (except Rukai) and a large number of languages in the Philippines, northern Borneo and northern Sulawesi are of the so-called ‘Philippine type’ (Himmelmann 2002, 2005, Ross 2002a). They have two voices: actor voice and undergoer voice. The latter is always transitive, and assumes two or three applicative-like forms which permit various semantic roles to assume subject position (this interpretation goes back at least to Starosta 1986). The transitivity of the actor voice remains a matter of controversy: it is morphosyntactically intransitive in some languages but apparently transitive in others, albeit of lower transitivity than the undergoer voice (Chang 2004, Liao 2004, Reid and Liao 2004, Ross and Teng 2005). Philippine-type languages typically have noun-phrase case-markers and case-marked pronoun sets indicating the

1 Reid (1982) suggests Amis as the closest Formosan relative of PMP, implying yet another Austronesian subgrouping, but I have shown in Ross (2005) that the innovations allegedly shared by Amis and PMP are either not exclusive or based on incorrect reconstruction.

522

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian grammatical function of the noun phrase or pronoun within the voice construction, and in many languages there are sets of nominative and genitive clitic pronouns which tend to occupy second (Wackernagel) position (Billings and Kaufman 2004). The organisation of this case-marking varies from language to language, and to describe it with an acceptable degree of faithfulness to the data means that cross- linguistically applicable case labels must needs be based on a listing of the grammatical functions found across the languages. The grammatical functions found in this analysis include first the core arguments of the verb:

(1) VSBJ subject of verbal clause AGT agent argument of undergoer-voice clause PAT patient argument of actor-voice clause, third core argument of undergoer-voice clause (e.g. patient when, say, a location or instru- ment is subject)

Usually, subject case-marking is identical in both verbal and non-verbal clauses. In this case I recognise a single SBJ grammatical function. However, sometimes more narrowly defined subject functions appear, and nonverbal clauses introduce functions of their own:

(2) SBJ subject of verbal and nonverbal clauses AVSBJ subject of actor-voice verbal clause NSBJ subject of nonverbal clause PREDN predicate noun

Certain peripheral arguments in verbal clauses need separate recognition:

(3) LOC location argument (‘at my place’, ‘from me’ etc), goal, source BEN beneficiary argument

Certain grammatical functions arise from information structure:

(4) DISJ disjunctive, i.e. one-word answer FRFOC focus-fronted argument TPC fronted topic, often followed by a topic marker

And finally possessor functions occur in noun phrases:

523

Malcolm Ross

(5) PSRA possessor adnominal (‘my’ etc) without a ligature PSRN possessor nominal (‘mine’ etc; in some languages it also occurs adnominally with a ligature)

On the basis of these functions I define a set of case-marker labels similar to those used by Huang et al. (1999).

(6) NEUT free form with functions including DISJ, TPC, FRFOC and one or more core grammatical functions (i.e. of SBJ, AGT and PAT) NOM free or clitic form serving as SBJ or VSBJ GEN free or clitic form serving as AGT and often as PSRA or occasionally PSRN PSR free form serving as PSRA and/or PSRN ACC free form serving as PAT only OBL free form serving as PAT and LOC (and sometimes in other peripheral functions) LOC free form serving as LOC (and sometimes in other peripheral functions)

Other abbreviations used in this article are: 1S etc first person singular etc; C common; D definite; EP exclusive plural; IP inclusive plural; P plural; PS personal; S singular; SP specific; V visible. Conventions used in the tables of data and reconstructions are:

(x) a form either with or without x occurred. [x] forms with and without x occur(red). (x,y) a form with either x or y occurred. [x,y] forms with both x and y occur(red). (*xxx) a parentheses around a reconstruction indicate uncertainty because the supporting data are insufficient.

2. Case-markers

Reconstructing the PAn pronoun paradigm presupposes a reconstruction of its case-marker paradigm, as the pronouns of Philippine-type languages incorporate case-marking and usually have the same case-marking possibilities as personal lexical NPs (common NPs often have somewhat different possibilities). Tables 1 and 2 set out the material for the reconstruction of PAn case-markers. This is not quite raw material. In

524

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian the interests of presentation, where data are available for several related dialects, I have reconstructed a local protolanguage (Proto Atayal from Atayal and Seediq dialects, Proto Amis, Proto Puyuma, Proto Rukai and Proto Bunun). The supporting data are in Appendix A. Reconstructing the PAn case-marking paradigm is not a simple task, and a glance at the tables gives some idea of the problems. Formosan case-markers tend to be (C)V monosyllables, where C- or ∅- indicates the case, -i a personal singular case-marker, -a a personal singular case-marker, and a vowel other than -i a common case-marker. This means that -a occurs in both personal plural and common case-markers (see discussion below). Thanks to diachronic shifts in constellations of grammatical functions, cognate forms in two Formosan languages quite often mark different cases. There is, for example, considerable crossover between OBL and LOC. Conversely, it is common for the same case in two languages to be differently marked. Some NOM forms are ∅-initial, others k-initial. As a result, cognate forms may appear in adjacent columns rather than underneath one another in the same column of Table 1 or Table 2, and non-cognate forms may occur in the same column.

Table 1: Common case-markers

NEUT NOM GEN ACC OBL LOC ligature Pazeh — ki ni — u di a 1 Saisiyat — ∅ no[ka] ka no BEN ray a P-Atayal +SP — *kuʔ, *na-kuʔ, — *ci-kuʔ, *iʔ, *suʔ — *kaʔ *na-kaʔ — *ci-kaʔ — — -SP — (*aʔ) *na — *c[i,u,a]ʔ — Thao — sa — — sa tu ? [y]a Faborlang — ya no — o de, i o 2 3 Siraya — ta na — ki tu ki 4 3 Tsou +D — ʔe, si, ta — — ta — ci 3 ±D, +SP, -V — ʔo — — to — 3 -D, -SP — na — — no, ne — 5 3 Kanakanavu — s[u]a, si — — s[u]a na — 6 3 Saaroa — a, ka — — na — — Paiwan — a nua, na — tua, ta, tu — a P-Rukai — *ka, *ku, *na — — (*sa) — — 3 P- Puyuma +D — *[i]na — — *kana *i — 3 -D — *a — — *ɖa i P-Amis *u *ku *nu *tu — *i [...-an] *a +D ? *iya *kiya *[nu]niya *tiya — P-Bunun — *a, (*ka), (*ca) — — (*i), (*ki)3 — (*a ), (*tu) 7 8 Kavalan — ∅/[y]a na — tu, ta ta ...[-an], [y]a sa9 PMP — *∅-/*k- *n- *t- *s- *d-, *i *a,* na

525

Malcolm Ross

1 Subject NPs also occur with ACC markers (Yeh 1991:37). 2 Siraya na apparently marks partitive, but not possessive or agentive. 3 GEN-OBL: AGT NP is OBL-marked but is coreferential with a GEN agreement marker. 4 ʔe ‘near speaker’, si ‘near hearer’, ta ‘near neither speaker nor hearer’. 5 No semantic distinction known (Tsuchida 1976:37). 6 Radetzky (2003) suggests that ka represents the grammaticisation of a demonstrative as a definite marker in all its contexts. 7 Kavalan a and ya are in free variation (Lee 1997:17). 8 ta occurs with human referents, tu with non-human (Li and Tsuchida 2006). 9 sa allative.

Table 2: Personal case-markers

NOM GEN ACC OBL LOC 1 Saisiyat ∅ ni hi ini kan, kala P-Atayal *iʔ ? *niʔ ?*iʔ ? *∅/*-an *kiʔ ? 2 Siraya ta ∅ — -aŋ — Paiwan S ti ni — cay — P tia nia — c(a)ia — P-Rukai *ki — — *-an — P-Puyuma S *i *ni — *ka-ni — P *na *na — *ka-na — P-Amis S *ci *ni *ci ...-an — (*i ci) P *ca *na *ca ...-an — ((*i ca)) Bunun =[k]at — — =[i]t — 3 Kavalan [y]a ti/∅ ti ni [tu] ti — [ta] ti PMP4 *si *ni — *ka ni, ka-y — There are no distinct personal case-markers in Pazeh, Favorlang, Tsou, Kanakanavu or Saaroa

1 Benefactives: the difference between these is not understood. 2 Siraya ta is NOM with both common and personal nouns. 3 Lee (1997) gives [tu] ti. Li and Tsuchida (2006) give the human marker as ta: they do not specify whether this occurs with personal nouns. 4 Forms reconstructed by Reid (1978).

526

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

In spite of all this, examination of the forms in Tables 1 and 2 suggests patterns. The distribution of initial consonants is not chance. Most obvious is that GEN markers usually begin with n-, a long recognised fact. On further examination more patterns emerge. Organising forms by their initial consonants gives the sets in Table 3. These include most of the forms in Tables 1 and 2, and some of the residue will be discussed below. The reflexes of some PAn consonants are complicated, and I have had to make decisions as to which correspondence set to assign certain forms to. Thus I have inferred that the s- of Thao sa in Table 3 reflects PAn *d-, but it could also reflect *Z-. The assignment of the set in the rightmost column to *d- is based on the fact that Puyuma ɖ- in ɖa only reflects *d-, not *Z-. Some of the other assignments are a little more arbitrary. The case-marking function of initial consonants is clear from Table 3, and the PAn forms that are apparently reflected in the data are given at the head of the table. The case labels of the reconstructions follow fairly obviously from the data, except for the assignment to NEUT of forms consisting of a vowel only. This assignment is based on two facts. First, many modern languages have a NEUT vs NOM distinction in pronouns, and it is likely that this distinction occurred in Proto Austronesian. Second, although only one modern Formosan language, Amis, maintains this distinction in its case-marking paradigm, I have assumed that the Amis distinction between *∅- NEUT and *k- NOM reflects a PAn distinction: this helps explain why both forms occur in the NOM paradigms of modern languages. Two languages which no longer reflect the NEUT/NOM distinction in their case-markers retain it in their personal pronouns, and the NEUT pronouns are marked by reflexes of PAn *i NEUT:PS:S: Pazeh i- and P-Puyuma *i- (see Appendices, §B.1 and §B.11.) Several forms are reflected in Tables 1 and 2 but are not included in the analysis in Table 3. The most important is PAn *si PS:NEUT, which is actually better supported than PAn *i PS:S:NEUT and *ki PS:S:NOM, reconstructed in Table 3. It is not clear to me what distinction there was between PAn *si and PAn *i or *ki.

2 (7) PAn *si PS:NEUT: Saisiyat hi PS:ACC, Paiwan ti PS:NOM, P-Amis *ci PS:NOM, PMP *si PS:NOM

There is also evidence that there was a PAn personal plural form *si-a, reﬂected by Amis ca and Paiwan tia (the *-a element is also reﬂected in the Puyuma plurals).

2 The fact that the Saisiyat reﬂex is ACC is inconsequential, as there has been a merger of NOM and ACC forms in Saisiyat, followed by loss of NOM forms.

527

Malcolm Ross

Easily confused with the above is PAn *Ci ‘proper-name marker’. Its reﬂexes do not mark case but follow a case-marker and indicate simply that a noun is a proper name.

(8) PAn (or Proto Pituish) *Ci ‘proper-name marker’: Thao θi- (in θiθu 3 NOM/ACC/GEN:3S), Siraya ti, Paiwan ti, Bunun -t ([k]a-t PS:NOM, =[i]-t PSR, ACC), Kavalan ti

Other forms not reconstructed in Table 3 are shown in (9), (10) and (11).

(9) PAn (or Proto Pituish) *i LOC (Blust 1995): P-Atayal *iʔ, Favorlang i, P-Puyuma *i, P-Amis *i, P-Bunun *i OBL, PMP *i LOC

Reﬂexes of PAn*i vary as to whether they precede another case-marker, but there are enough lan-guages where i occurs without a further case-marker for us to assume that this was the situation in Proto Austronesian.

(10) PAn *ka OBL: Saisiyat kan, kala BENEFACTIVE, P-Puyuma *ka-na OBL:+SP, ka-ni PS:OBL:S, ka-na PS:OBL:P, PMP *ka ni, *ka-y

This PAn *ka was distinct from NOM *ka, and preceded a case-marker—on the data in (10) this was GEN. PAn *ka was thus a preposition and marked a beneficiary or other peripheral role.

(11) PAn (or Proto Pituish) *-an LOC: P-Atayal *-an PS:OBL, Siraya -an PS:OBL, P-Amis *-an LOC, PS:ACC, P-Rukai *-anə PS:OBL, Kavalan -an LOC

PAn *-an appears to have been a LOC suffix and is considered in §3.2.1.

3 The Paiwan form is ti instead of expected *tsi, and this probably reﬂects a formal conﬂation with Paiwan ti PS:NOM. These are clearly homophonous but distinct morphemes, as Paiwan ti ‘proper-name marker’ does not mark case but precedes a personal noun that is head of a noun phrase containing one or more attributive items. Such a noun phrase is marked with a common case-marker (data provided by Anna Hsiou-chuan Chang).

528

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

Table 3: A provisional reconstruction of Proto Austronesian case-markers

PAn *∅- NEUT *k- NOM *n- GEN *C- ACC *s- ? OBL *d- ? LOC C:? ? *[y]a *k-a *n-a *C-a *s-a *d-a C:? ? *u *k-u *n-u *C-u *s-u ... PS:S *i *k-i *n-i *C-i — — PS:P ...... *n-i-a ... — — Pazeh ... ki NOM ni GEN ...... di LOC Saisiyat C ... ka ACC, NOM no GEN, OBL ...... ray LOC PS ...... ni GEN ...... P-Atayal C:-SP *aʔ NOM ... *naʔ GEN *caʔ OBL ...... C:+SP ... *kaʔ NOM *na-kaʔ GEN *ci-kaʔ OBL ...... *kuʔ NOM *na-kuʔ GEN *ci-kuʔ OBL ...... PS *iʔ NOM ... *niʔ GEN ...... Thao ...... tu LOC ? sa NOM, OBL Favorlang ya NOM ... no GEN ...... Siraya C ...... na partitive ... tu LOC ... Tsou C:+D ... ʔe NOM ...... C:+D:-V ... ʔo NOM ...... Kanavu ...... na LOC ... s[u]a NOM, OBL ... Saaroa a NOM ...... Paiwan C a NOM ... n[u]a GEN ... t[u]a OBL ... PS:S ...... ni GEN ...... PS:P ...... nia GEN ...... P-Rukai C ... *ka NOM ...... *ku NOM ...... PS:S ... *ki NOM ...... P-Puyuma C:+D ...... C:-D *a NOM ...... *ɖa OBL PS:S *i NOM ... *ni GEN ...... P-Amis C *u NEUT *ku NOM *nu GEN *tu ACC ...... PS:S ...... *ni GEN ...... PS:P ...... *na GEN ...... P-Bunun ... *ka NOM ... *tu ligature ...... Kavalan C [y]a NOM ... na GEN ta ACC, LOC sa allative ...... tu ACC ...... PS ...... ni GEN ...... PMP *∅-/*k- NOM — *n- GEN *t- ACC *s- OBL *d- LOC

529

Malcolm Ross

The form of the PAn ligature, *a, is self-evident on the basis of the data in Table 1. Two caveats apply to the reconstructed ‘paradigm’ in Table 3. First, if Sagart’s subgrouping of Austronesian is correct, then *∅- NEUT, *C- ACC and *s- OBL (?) are reconstructable only in his Proto Pituish, as they are not reﬂected in Pazeh or Saisiyat. Second, no modern language has a system as complex (or as symmetrical) as the reconstructions in Table 3, and I think it would be wrong to infer that all the reconstructions in the table existed in Proto Austronesian simply because they are reﬂected in modern languages. Instead, I think it is more likely that what existed in Proto Austronesian was perhaps not a fixed paradigm like the one reconstructed in Table 3, but a set of determiners, possibly *a and *u, both ‘common’, and *i ‘personal’ to which were prefixed the case-marking morphemes *k- NOM, *n- GEN, *C- ACC, and so on (the hyphens in the reconstructions are intended as a reminder of this). We see such case-marker + determiner sequences arising cyclically in the histories of Philippine-type languages as a case-marker suffers attrition of its earlier function, and its form is reanalysed as a caseless determiner. As a result, what looks like double case-marking arises. P-Atayal *na-kuʔ GEN:+SP and *ci-kuʔ OBL:+P are cases in point. Each was formed historically from *na- GEN or *Ci- ACC plus what was in P-Atayal (and still is in Mayrinax) the specific nominative form *kuʔ. Saisiyat no-ka GEN (< GEN + ACC/ NOM), Tamalakaw Puyuma ni-nina GEN:-D:+SP (< GEN + GEN) and Sakizaya Amis nu-niya GEN:+SP (< GEN + GEN) are outcomes of parallel processes, and I infer that the PAn forms were also case-marker+ determiner sequences. The observation that the vowel *-i is characteristic of personal case-markers is not new. The functions of the *-a/*-u distinction among PAn common case-markers are much less clear, however. In Ross (2002a) I mentioned an unpublished analysis to the effect that the vowel distinctions encoded a present vs absent distinction, but Blust (2005) correctly points out that this inference is not well supported by the data. Curiously, however, he presents an alternative interpretation which is just as problematic. Examining only n-initial GEN forms, he concludes that *nu was the common GEN marker, *ni the personal singular GEN marker, and *na the personal plural GEN marker. I agree with him regarding the functions of *nu and *ni, but the forms in Table 3 are sufficient to cast doubt on his characterisation of *na. Genitive case-markers of the form na have two sources in Formosan and Philippine languages. First, PAn *na stood alongside *nu as a common GEN marker, as indicated by Proto Atayal *naʔ, Siraya na, Paiwan na, Kavalan na in Table 3 and by similar forms in Philippine languages (Reid 1978). Second, Amis na marks personal GEN plural, and to this Blust adds Southern Bikol na, also personal GEN plural. However, the presence in Table 3 of Paiwan nia with the same function suggests that the protoform was not *na but *ni-a PS:P:GEN, incorporating the same plural marker

530

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

*-a as we encountered above in *si-a PS:P:NOM. The evidence of paradigmaticity is important here. The well supported reconstruction of the PAn *a C:NOM supports the reconstruction of *na C:GEN, and the reconstruction of *si-a (and the presence of plural -a in Paiwan cai-a PS:P:OBL as opposed to cai PS:S:OBL) supports the reconstruction of *ni-a. The derivation of na from *nia is straightforward: *nia>*ña> na. On typological grounds, one might expect the *-a/*-u distinction among PAn common case-markers to encode a ±definite or a ±specific distinction. There is miniscule evidence for such a distinction in the Atayal (Appendices, §A.1) and Rukai data (§A.2), but it is too scant to build a case on, and so the puzzle remains. Starosta (1992) infers that the *-i of personal case-markers was originally the *-i of definite case-markers. This is possible, but lacks empirical support, as the vowel *-i is reﬂected in the common case-markers of only one Formosan language, Pazeh, and this has lost the common/personal distinction. Whether *si-a PS:P:NOM and *ni-a PS:P:GEN are reconstructable to PAn depends on one’s subgrouping assumptions. On the basis of Blust’s subgrouping, they are. On the basis of Sagart’s, they are not. Sagart (2004) divides his large Pituish subgroup into a collection of languages which includes Atayal-Seediq, Thao and certain extinct languages (it corresponds to Blust’s Western Plains group) and an ‘Enemish’ subgroup which comprises most of Austronesian (Siraya, Tsou, Kanakanavu, Saaroa, Paiwan, Rukai, Puyuma, Amis, Bunun, Kavalan and Malayo-Polynesian). On the basis of the data in Table 3, *si-a and *ni-a are reconstructable to Proto Enemish but not to PAn.

3. Personal pronouns 3.1 Introduction

Table 4 brings together the PAn personal pronouns reconstructed below in §3.2 and §3.3. Forms whose reconstruction is not firmly supported by the data are shown in parentheses. Reconstructions of first and second person pronouns are presented in §3.2 and of third person pronouns in §3.3 (no 3P pronouns are reconstructable). First and second persons and the third person are treated separately because their histories are rather different.

531

Malcolm Ross

Table 4: A preliminary reconstruction of Proto Austronesian personal pronouns

1S 2S 3S 1IP 1EP 2P NEUT i-aku iSu[qu] s-ia ita i-ami i-mu[qu], i-amu NOM1 aku Su[qu] ia (i)ta ami mu[qu], (amu) NOM2 =ku, =Su (-ya) =(i)ta =mi[a], =mu =[S]aku =[S]ami ACC i-ak-ən iSu[qu]-n ... ita-ən i-ami-n i-mu[qu]-n GEN1 =[a]ku =Su (-ya) =(i)ta =mi[a] =mu GEN2 (=)m-aku (=)m-iSu ... (=)m-ita ((=)m-ami) (=)m-amu GEN3 n-aku n-iSu n-ia ni-ta ni-am ni-mu n-ami n-amu

It is immediately obvious that too many sets of pronouns are reconstructed in Table 4. There are two NOM sets and three GEN sets, and this seems a little implausible. I return to this matter in §3.6. The raw material for the reconstructions is set out in Appendix B. It consists of pronominal paradigms for PMP and from all Formosan languages, and includes the reconstructions for local protolanguages that are used in the body of the paper. In the tables in §3.2 and §3.3 only the supporting data immediately relevant to the reconstruction are given. Pronoun forms that do not directly reﬂect a reconstructed form are omitted. The proportion of the data in Appendix B that is omitted from the tables is strikingly high, and §3.4 discusses why this should be.

3.2 Reconstructing first and second person Proto Austronesian pronouns

The data on which the reconstructions of first and second person pronouns in Table 4 are based are laid out in Table 5. There is no strong basis for reconstructing PAn *(=)m-ami GEN2:1EP as it is reﬂected only in Malayo-Polynesian languages. However, if an m-initial form occurred in PAn, this is the form we would expect it to have. Wulai Atayal, Paran Seediq and Siraya =mian looks like an m-initial 1EP form, but it appears to be innovatory, perhaps a metathesis of PAn *ni-am GEN3:1EP under analogical pressure from m-initial *(=)maku GEN2:1S.

532

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

Table 5: Towards a reconstruction of first and second person Proto Austronesian pronouns

1S 2S 1IP 1EP 2P PAn NEUT *i-aku *iSu[qu] *ita *i-ami *i-mu[qu], *i-amu Pazeh NEUT y-aku i-siw i-ta y-ami i-mu Saisiyat NOM ya[k]o ... ʔitaʔ yami ... P-Atayal NEUT (*i-aku) *isuʔ *itaʔ (*i-ami) (*i-amu) Thao NOM1 yaku ihu ita yami-n ... Siraya NEUT ĭau ...... P-Bunun NEUT *ðaku *suʔu *ita *ðami *mu[ʔu] PMP NEUT *i-aku ...... PAn NOM1 *aku *Su[qu] (*i)ta *ami *mu[qu], (*amu) Pazeh NOM aku siw ta ami mu Saisiyat NOM ... ʃoʔo ...... moyo Tsou NEUT aʔo suu ...... muu P-Rukai NEUT *aku ...... 1 P-Amis NEUT *aku *isu *ita *ami *amu PMP NOM1 *aku ...... PAn NOM2 *=ku, *=Su *=(i)ta *=mi[a], *=mu *=[S]aku *=[S]ami P-Atayal NOM *=[ca]ku *=suʔ *=taʔ (*=cami) (*=cimu) Thao NOM2 wak — — — — Siraya NOM =koh ...... Tsou NOM -ʔo/-ʔu -su -to -mza -mu Kanavu NOM =ku ...... Saaroa NOM =aku =u =ita ... =mu P-Rukai NOM *ku *su *ta ... *=mu P-Puyuma NOM *=ku *=(y)u *=ta *=mi *=mu P-Bunun NOM2 *=[s]a-k *=a-s *=[a-]ta *=[s]a-m *=a-mu Kavalan NOM =iku =su =ita =imi =imu PMP NOM2 *=aku ... *=ta ...... PAn ACC *i-ak-ən *iSu[qu]-n *ita-ən *i-am-ən *i-mu[qu]-n Saisiyat ACC yak-in ʔi-ʃoʔo-n ...... Thao ACC yak-in ihu-n ita-n [y]amin ... Paiwan NOM =[a]ken =[e]sun =[i]cen =[a]men =[e]mun P-Bunun ACC *ðak-un *suʔu-n *it-un *ðam-un *muʔu-n PMP PSR *[y]akən ... *[y]atən *[y]amən ...

533

Malcolm Ross

PAn GEN1 *=[a]ku *=Su *=(i)ta *=mi[a] *=mu Pazeh GEN1 — — ta- — — P-Atayal GEN *=ku, *=suʔ *=taʔ ...... Thao GEN2 -[a]k, -[i]k — — — — Siraya GEN =au =uhu =ǐtta ...... Tsou GEN -ʔo/-ʔu -su -to -mza -mu Kanavu GEN1 =aku =su =ta =mia =mu Saaroa GEN =ku =u =ta ... =mu 3 3 Paiwan GEN ku= su= ca= nia= nu= P-Rukai PSR ... *=su *=ta ... *=mu P-Puyuma PSR1 ... *=u (*=ta) (*=mi) (*=mu) GEN *ku= ... *ta= *mi= *mu= P-Amis GEN1 =aku =isu =ita — =amu 2 2 P-Bunun GEN *=ku *=su[ʔu] *=ta ... *=mu[ʔu] Kavalan GEN -ku -su -ta ...... PMP GEN1 *=ku *=mu *=ta *=mi ... PAn GEN2 *(=)m-aku *(=)m-iSu *(=)m-ita (*(=)m-ami) *(=)m-amu Saisiyat PSR1 ...... m-itaʔ ...... P-Atayal GEN *=m-aku ...... (*=m-amu) Thao GEN1 ... m-ihu m-ita ...... Siraya GEN =mau =moɣo =mita =mian =momi Kanavu GEN1 =maku =musu =mita ...... P-Rukai NEUT ... *musu *mita ... *mumu P-Amis GEN2 m-aku m-isu m-ita ...... P-Bunun PSR ...... *=mita ...... PMP GEN2 ...... — *=mami ... PAn GEN3 *n-aku *n-iSu *ni-ta *ni-am *ni-mu *n-ami *n-amu Pazeh GEN2 n-aki ni-siw ni-ta ni-am ni-mu 4 Saisiyat PSR1 ... ni-ʃoʔ ... ni-yaʔom ni-mo-n P-Atayal GEN ...... *=ni-am (*=n-amu) Thao GEN1 nak ...... nam ... Kanavu GEN2 =naku =nsu =nta =nmia P-Rukai PSR ...... *=na[m]i *nimu NEUT *naku ...... *na[m]i *nimu P-Puyuma PSR2 ...... *niam= ... P-Amis GEN2 (n-aku) (n-isu) ... niam n-amu P-Bunun PSR *=nak ... *=nita *=nam *=nuʔu

534

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

Kavalan GEN ...... -niaq -numi PMP GEN2 *=n(a)ku *=nihu — *=nami ...

1 Sakizaya Amis has kita NEUT:1IP where Nataoran and Central Amis have ita. There is a reasonable possibility that Sakizaya kita is copied from kita NOM:1IP, where k- reflects the NOM case-marker and is an innovation limited to Amis. 2 P-Bunun *=su[ʔu] and *=mu[ʔu] suggest at first sight the reconstruction of PAn *Su[qu] and PAn *mu[qu]. However, P-Bunun *=ðami GEN:1EP clearly reflects PAn *=i-ami NEUT:1EP, and it is possible that the P-Bunun *=suʔu and *=mu[ʔu] also reflect the transfer of forms from the NEUT paradigm. 3 Paiwan nia= GEN:1EP and nu= GEN:2P appear to reflect PAn *=mia and *=mu with an unexplained *m-> n- substitution. 4 -n of ni-mo-n PSR1:2P copied from ʔini-mo-n ACC:2P.

3.2.1 Should Proto Austronesian oblique or locative pronouns be reconstructed?

It is tempting also to reconstruct a PAn oblique/locative personal pronoun set. In Formosan languages these sets, shown in Table 6, take two forms. In Saisiyat and Proto Puyuma, they consist of kan plus a pronominal base, but the pronominal bases in the two languages do not do not correspond with one another. In Saisiyat the base is the NOM form, except for 1S, where it is the PSR. In Proto Puyuma the base is the one to which i- PS:S is prefixed to form NEUT pronouns: this base was itself originally copied from a GEN form in *n-. From this difference between the two languages we may infer that the kan- pronouns were innovated independently in each language. The PAn oblique marker *ka, reconstructed in §2, presumably occurred with personal noun phrases, and these included NEUT and/or GEN pronouns. This construction was the starting point for the innovations in Saisiyat and Proto Puyuma. In Pazeh, Proto Atayal, Siraya, Kanakanavu, Proto Rukai, Proto Amis, Proto Bunun and Kavalan, oblique/locative pronouns are formed with reflexes of the PAn suffix *-an. Here, too, we are confronted by a reconstructive problem, as the bases to which the reflex of *-an is attached vary from language to language and do not correspond with one another. In Proto Atayal the bases resemble the NOM enclitics. In Pazeh and Proto Bunun, the base is the NEUT form, reflecting the PAn NEUT form. In the other languages the base is either the NEUT or the NOM form, but this has undergone innovation relative to PAn. Again, the obvious inference is that the *-an construction occurred in PAn, but morphologised pronouns with *-an did not. The *-an construction consisted of a noun suffixed with *-an and preceded by a locative preposition. The noun was apparently either common or personal, including personal pronouns. The construction survives with common nouns in Amis and Kavalan (see Table 1). The identity of the preposition is not

535

Malcolm Ross recoverable (Pazeh reﬂects *di, Proto Amis and Proto Bunun *i). Proto Amis and Kavalan reﬂect a prefix *Ca- on the pronoun, apparently PAn *Ca ACC (see Table 3).

Table 6: Oblique/locative personal pronouns in Formosan languages

1S 2S 1IP 1EP 2P Saisiyat LOC kan-man kan-ʃoʔ kan-ʔitaʔ kan-yami kan-moyo P-Puyuma OBL *kan-(iŋ)ku *kan-nu *kan-ta *ka-niam *kan-mu Pazeh LOC yaku-[n]an i-siw-an i-ta-an y-ami-[n]an i-mu-[n]an P-Atayal OBL *ke-nan *su-nan *ita-[na]n (*ca)mi-nan (*ca)mu-nan Siraya OBL ǐau-an ǐmhu-an ǐmǐtta-n ǐmian-an ǐmumi-an Kanavu GEN -OBL ʔiku-an kasu-an kita-nan kimi-an kamu-an 1 2 1 P-Rukai OBL *naku-a[nə] *su-a[nə], *mita-a[nə] *na[m]i-a[nə] *mu-a[nə], *musu-a[nə]2 *[ni]mu-a[nə] P-Amis OBL *[i]t-aku-an *[i]t-isu-an *[i]t-ita-an, *[i]t-aman *[i]t-amu-an *[i]kita-an P-Bunun LOC *ðaku-an *suʔu-an *ita-an *ðami-an *muʔu-an Kavalan LOC ta-ma-iku-an ta-ma-isu-an ta-ma-ita-an ta-ma-imi-an ta-ma-imu-an

1 Base is the n-initial GEN form. 2 Base is the m-initial GEN form.

3.3 Reconstructing third-person Proto Austronesian pronouns

Third-person personal pronouns have a rather different history from non-third- person. It is clear from forms occurring in Pazeh, Tsou, Amis, Bunun and Kavalan, and from the non-correspndence with one another of third-person forms in general, that third-person personal pronouns in Formosan languages either are derived from or still are demonstrative pronouns. The reconstruction of demonstratives demands a paper to itself, and I shall not undertake it here. The few third-person forms that appear to be cognate are shown in Table 7. Sorting them out is tricky, because forms that appear superficially to be cognate in fact are not. The problem is similar to that encountered in regard to PAn *si PS:S:NOM and *Ci ‘proper-name marker’ in (7) and (8) above. Two PAn base forms are reflected, namely *Cia and *sia, and these are shown against their reflexes in the table. PAn *Cia is, quite simply, a base, and perhaps a demonstrative base, as the Pazeh reflex, sia, is a member of a three-member demonstrative set distinguishing CLOSE, DISTANT and INVISIBLE: sia is the INVISIBLE member. Note that *ni PS:GEN is preposed to reflexes of *Cia, whereas it replaces the first syllable of *sia. This implies that the first syllable, *si, is PAn *si PS:

536

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

NEUT, an inference which receives support from PMP reconstructions where we find (i) both *sia and *ia as NOM forms (and enclitic *=ya), implying that the base was *ia and that *sia< *si+ *ia; (ii) plural *sida and enclitic *=da, implying that the plural base was *da and that *sida< *si+ *da. The conclusion to which the previous paragraph leads is that we have weak evidence for

(12) PAn (or Proto Pituish) *s-ia NEUT:3S, (*ia NOM:3S), (*=ya NOM:3S/ GEN:3S), *n-ia GEN:3S

Pazeh and Saisiyat 3P forms reﬂect a plural marker *-a following the case-marker, a form which I discussed in §2. However, Saisiyat and P-Atayal *-la- also apparently mark the plural, but remain otherwise unconnected with other data. The PMP plural base *da has no known Formosan cognate.4

Table 7: Towards a reconstruction of Proto Austronesian third-person personal pronouns

PAn base NEUT NOM1 NOM2 GEN1 GEN2 3S Pazeh *Cia i-sia sia — — ni-sia Saisiyat *Cia hi-sia, ʔi-sia sia — — ni-sia P-Atayal *sia *hiyaʔ — — — *=niya, *=na PMP *sia — *sia, *ia *=ya *=ya *=nia 3P Pazeh *Cia y-a-sia a-sia — — n-a-sia Saisiyat *Cia hi-la-sia la-sia — — n-a-sia P-Atayal *sia *re-hiyaʔ — — — (*=laha) PMP *sida — *sida *=da *=da (*=nida)

3.4 Why modern forms often do not reﬂect Proto Austronesian forms

There are a number of reasons why modern forms do not directly reﬂect PAn forms. It is perhaps typical of pronoun paradigms the world over that forms are often altered or replaced on the basis of analogies which render the contrasts between case-forms more transparent (i.e. regular) or which do the same for differences among persons. A fairly

4 After these observations on *-a and *-la- were written, I discovered that Elizabeth Zeitoun has argued for the reconstructability of these markers in various public presentations since 2001. The most extensive written account is Zeitoun (2006), the scope of which is wider than my brief notes here.

537

Malcolm Ross common outcome of analogical pressure is that a set of pronouns with a particular case-marking function is reanalysed as a set of bases for the formation of a further pronoun set. For example, the non-third-person NEUT forms in now extinct Siraya were ĭau 1S, ĭ-mhu 2S, ĭ-mĭtta 1IP, ĭ-mian 1EP, ĭ-mumi 1IP. The first of these, ĭau, directly reflected PAn *iaku NEUT:1S (see Table 4), but the others reflected the Siraya m-initial GEN forms =muhu, =mĭtta, =mian, =mumi with preposed i-. The GEN forms, reflecting PAn GEN *miSu 2S, *mita 1IP, *mian 1EP, *mamu 2P, had thus been reanalysed as pronominal bases for the innovatory NEUT forms (and for PSR forms in ā-: see Appendices, §B.13). The i- probably reflected PAn *i PS:S, although it no longer functioned as a case-marker in the Siraya manuscripts. The innovatory NEUT forms functioned in turn as bases for the Siraya OBL forms ĭau-an, ĭmhu-an, ĭmĭtta-n, ĭmian-an, ĭmumi-an. Another case in point is Paiwan in (13), which reflects a replacement of all PAn forms except the PAn GEN enclitics. The first major change that must have occurred in Paiwan is that reflexes of the PAn ACC forms in Table 4 became the NEUT set, becoming enclitic nominatives in Paiwan (NOM in 13). New forms for the other cases were then constructed with case-markers. The PAn person marker *si (cf 8) came to mark the NEUT set, the GEN personal case-marker *ni the PSR set. I am not sure of the origin of the OBL set.

(13) 1S 2S 1IP 1EP 2P NEUT ti-aken ti-sun ti-cen ti-amen ti-mun NOM =(a)ken =(e)sun =(i)cen =(a)men =(e)mun GEN ku= su= ca= nia= nu= PSR ni-aken ni-sun n-icen ni-amen ni-mun OBL canu-aken canu-sun canu-icen canu-amen canu-mun

An examination of the data in Appendix B shows a number of cases in which a new NEUT or NOM set has been formed by preposing a case-marker to an existing pronoun set. A transparent instance appears in Amis (B.12), where the old NOM set has become the NEUT set, and a new NOM set has been created by preposing the NOM common case-marker to the new NEUT set. Roughly parallel processes have occurred in Bunun (B.13) and Kavalan (B.14). In Proto Puyuma a new NEUT set was created by preposing *i to the GEN clitics (B.11). There is also a tendency for core and possessor forms to become clitics (and then affixes). Thus PMP GEN enclitics in *n- are derived from an earlier free set formed with the PAn case-maker *ni GEN:PS:S. A change which affects GEN clitics in their AGT function has been noted by Starosta et al. (1982) and Wolff (1996). The processes which they reconstruct differ a little, but their essence is that a PAn agent GEN enclitic was attached to the first item of the

538

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian clause—either a verb or an auxiliary. When the enclitic was attached to an auxiliary, it accordingly preceded the verb. In various Formosan and Philippine languages the auxiliary has been lost, leaving the GEN clitic as proclitic to the following verb. Among Formosan languages, this has occurred in Paiwan (see third row of 13), in Puyuma (Appendices, B.11) and apparently in Pazeh, where just ta- GEN:1IP survives. A number of changes of the kinds described above have occurred in the Rukai dialects, but here the processes are complicated by the fact that the ergative-like alignment of other Formosan languages has been replaced by an accusative alignment. Tracking the history of Rukai pronouns requires a paper to itself.

3.5 Did PAn have polite pronoun forms?

In (14) Blust’s (1977) nominative forms and my neutral forms from Table 4 are compared.

(14) 1S 2S 3S 1IP 1EP 2P 3P Blust’s nominative *i-aku *i-Su, *si-ia *i-kita *i-kami *i-kamu *si-iDa *i-ka-Su Ross’ neutral *i-aku *iSu[qu] *s-ia *ita *i-ami *i-mu[qu], … *i-amu

Apart for minor formal differences, the two sets differ only in Blust’s reconstruction of forms with *k-. He reconstructs *i-ka-Su 2S as a polite alternant to *i-Su 2S. He also reconstructs plural forms with *k- where I have no *k-. On the other hand, he reconstructs no *k-less plural forms. That these are in fact reconstructable is shown clearly in Table 5. The question is: are the forms with *k- reconstructable for PAn? The relevant data are set out in Table 8. The level to which the forms with *k- are reconstructable again depends on one’s subgrouping assumptions. On Sagart’s (2004) assumptions, they are re-constructable to Proto Enemish but not to PAn. I will not enter into the subgrouping controversy here. What matters is that the contrast between *k-less forms and forms with *k-, shown in (15), is reconstructable, be it PAn or somewhat later.

(15) 2S 1IP 1EP 2P k-less: NEUT *iSu[qu] *ita *i-ami *i-mu[qu], *i-amu with*k-: NEUT *i-ka-Su *i-kita *i-kami *i-kamu

539

Malcolm Ross

Harvey (1982) also reconstructs this contrast, but declines to attribute it to politeness on the grounds that such a contrast subsists in no modern language. The fact that the contrast does not occur in 1S, 3S or 3P, however, is circumstantial evidence that this is a politeness contrast and that Blust’s ‘first politeness shift’, which introduced *kaSu, also introduced the other forms with *k- in (15).

Table 8: Reconstructing first and second person pronouns with *k-

2S 1IP 1EP 2P PAn (?) NEUT *i-ka-Su *i-k-ita *i-k-ami *i-k-amu Kanavu NEUT ii-kasu ii-kita ii-kimi ii-kamu PMP NEUT *ikahu *i-kita *i-kami *i-kamu NOM1 *ikahu *kita *kami *kamu PAn (?) NEUT *=ka-Su *=k-ita *i-k-ami *i-k-amu 1 Siraya NEUT =kow =kǐtta =kame =kamu Kanavu NEUT =kasu =kita =kimi =kamu PMP NOM1 *=ka(hu) *=ta *=kami *=kamu

1 The expected form is *=kahu, so =kow reflects loss of -h-.

3.6 Too many pronoun sets?

I observed above that the data require us to reconstruct implausibly many PAn pronoun sets: two NOM sets and three GEN sets. Why should this be? The most likely explanation is that some forms in the modern languages reflect developments which took place at various times after the break-up of PAn. If Sagart’s subgrouping is correct, then the NOM enclitics are reconstructable only to Proto Pituish, as they are not reflected in Pazeh or Saisiyat. This may mean that the encliticisation of NOM forms had not occurred in PAn and occurred only after the ancestor of the Pituish subgroup had separated from the ancestors of Pazeh or Saisiyat. Indeed, NOM encliticisation may have occurred independently in different languages, as we find the *k- forms encliticised in Siraya and Kanakanavu (Table 8) and *k-less forms in other languages (Table 5). Encliticisation of GEN forms did occur in PAn, however, and one consequence of NOM encliticisation was that there were now two sets of enclitics, NOM and GEN (Table 4) which were very similar in form. Probably ambiguities arose in certain contexts, and speakers would naturally seek means of disambiguation. This seems to have been achieved by the use of two other GEN sets: m-initial and n-initial. Of the two, it seems probable that the m-initial set was earlier, as it is reconstructable to PAn (*=m-ita is reflected in Saisiyat) and we cannot readily

540

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian identify the *m- element. On the face of it, the n-initial set is also reconstructable to PAn, but we can identify the *n- element as the case-marker *ni PS:GEN (Table 3), which had a separate identity in PAn. It marked a GEN personal noun phrase, and we may infer that this personal noun phrase could also be an unmarked pronoun. On this inference, PAn *ni aku GEN + 1S, for example, was a phrase, not a word, and PAn *n-aku was an elided form of the phrase which was independently grammaticised in various daughter-languages. That later grammaticisations occurred is evidenced by Paiwan, where the forms ni-aken PSR:1S and so on (shown in 13) must have arisen after the erstwhile ACC form in *-ən became the Paiwan base form. The fact that the pairs of alternants *ni-am/*n-ami GEN: 1EP and *ni-mu/*n-amu GEN:1EP are reconstructable is quite possibly an artefact of different and independent grammaticisations of PAn*ni ami and*ni amu. It is noteworthy that there are more reﬂexes of PAn *m-aku GEN:1S, *m-iSu GEN:2S and *m-ita GEN:1IP in Table 5 than there are of *m-ami GEN:1EP and *m-amu GEN:2P, yet the distribution of n-initial reﬂexes is the other way round. It is possible either that *m-ami and *m-amu were reduced to *mi and *mu by haplology or that forms with a sequence of two *m- onsets were avoided in favour of analytical forms with *ni. The upshot of this section is that Table 4 may indeed include too many PAn pronoun sets, and that the rows marked NOM2 and GEN3 should perhaps be removed from it, as the enclitic NOM and *n-initial GEN sets may be artefacts of later developments in various daughter-languages.

3.7 Proto Malayo-Polynesian innovations

The PAn pronominal system shown in Table 4 evolved into the PMP system in Table 9. Changes include what Blust (1977) calls the second politeness shift, a set of innovations that defines the Malayo-Polynesian subgroup. Its elements (in terms of the reconstructions made here) are listed in (16).

(16) a. The PAn plain neutral *i-Su 2S is lost, and PMP *ikahu, reflex of PAn polite *i-ka-Su, becomes the default neutral 2S pronoun. b. PMP *=mu GEN:2S reflects the PAn clitic *=mu 2P, and the PAn clitic *=Su 2S is lost (although the long clitic *=nihu, reflex of PAn *(=)ni-Su, continues). c. PMP has new additional 2P forms, neutral *[i]ka-ihu and *kamu-ihu and genitive *=ihu, *=nihu, *=mu-ihu which incorporate *-ihu, apparently reflecting PAn neutral *i-Su 2S.

The relevant forms are bolded in Table 9.

541

Malcolm Ross

Table 9: A very tentative reconstruction of Proto Malayo-Polynesian pronominal forms

1S 2S 3S 1IP 1EP 2P 3P NEUT i-aku ikahu siya i-kita, ita i-kami i-kamu i-ka-ihu, — kamu-ihu NOM1 aku (i)kahu iya kita, i-ta kami kamu ka-ihu sida NOM2 =(h)aku =ka(hu) =∅, =ta =kami =kamu =ka-ihu, =da =ya =kamu-ihu GEN1 =ku =mu =ya =ta =mi — =ihu, =da -mu-ihu GEN2 =n(a)ku =nihu =niya — =mami, — =nihu =nida =nami PSR [y]akən imu, ihu — [y]atən [y]amən — ihu, inihu, — imu-ihu

The reconstructions presented here not only confirm that Blust was right in positing his PMP innovations, but that PMP underwent further innovations in its pronoun paradigms.

(17) a. The PMP NEUT and NOM forms *ikahu 2S, *[i]kita 1IP, *[i]kami 1EP, and *[i]kamu 2P all reﬂect the polite PAn NEUT forms *[i]ka-Su, *[i]k-ita, *[i-]k-ami and *[i-]k-amu, whilst the plain PAn NEUT forms *[i-]Su, *ita, *i-ami and *i-amu are lost (this is a widening of Blust’s 16a). b. The PMP 3P forms have no Formosan cognates, and appear to be a PMP innovation. c. The distinct (and incomplete) PMP NOM clitic set was created by cliticising free forms, leaving the old short clitic set to serve only as short genitives in PMP. This innovation evidently postdates the second politeness shift, as the 2S clitic form is *=kahu (< PAn polite *ka-Su, not plain *Su). d. PAn *m- clitics have disappeared, except for *=mami 1EP, and have otherwise been replaced by PAn *n-, whose members are now clitics. e. PAn ACC forms became PMP PSR forms. f. PAn oblique forms in *-an are lost.

542

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

4 Conclusions

The PAn reconstructions in Table 4 are very similar to those in Ross (2002b:36). Differences resulting from a re-examination of the data are minor. It is no longer clear to me whether polite forms in *k- are reconstructable to PAn (§3.5). I no longer reconstruct 3P forms (§3.3). For 1EP I have added the alternant *=nimu GEN:1EP and *=ni-mu GEN:2P forms for reasons that are obvious from Table 5, but these forms are probably not reconstructable to PAn (§3.6). I have added ACC forms (these were only mentioned in the text in 2002). The OBL forms are harder to reconstruct, as the locative/oblique marker *-an has continued to be productive in various Formosan languages, and has continued to generate fresh oblique forms (§3.2.1). A small detail is the reconstruction of the alternant second-person forms *-Suqu 2S and *-muqu 2P (the latter reﬂected only in Bunun and therefore very doubtful). These perhaps represent lengthenings of the basic form for vocative emphasis. In 2002 I reconstructed only one set of monosyllabic enclitics, which served as both NOM and GEN (as Starosta et al. 1981 point out). I show them as two sets in Table 4, but recognise that the NOM set probably did not yet occur in PAn (§3.6). Both the 2002 reconstruction and the version presented in Table 4 differ substantially from previous reconstructions, because they take the full range of available data and attempt a reconstruction based on paradigmaticity as well as form. Dahl (1973) reconstructs a free set (corresponding to my free NOM set, and incorporating the *∅-/*k- alternation which I have attributed to politeness) and a clitic set (corresponding to my monosyllabic enclitics). Blust (1977) reconstructs only two sets, a nominative and a genitive, and points to the existence of what he takes to be a possessive nominal (PSRN) set: it corresponds in form to my accusative set. His nominative set corresponds to my neutral set. His genitive set corresponds to none of my sets. He posits two genitive variants, one with *i- and one with *ni- (e.g. *i-ku, *ni-ku GEN:1S). The arguments for these are too complex to rehearse here, but they are based almost entirely on Malayo-Polynesian data and represent developments that have occurred in Malayo-Polynesian languages and therefore postdate PAn. There is certainly no Formosan evidence for the *i- set, as Harvey (1982:75) points out. The *ni- set is suggestive of post-PAn developments like those described brieﬂy in §3.6. It must be said, however, that Blust states explicitly that he is not about a fullscale reconstruction of the the PAn pronoun paradigm. Instead he is concerned to show how case-marking shaped early Austronesian pronoun sets and particularly how a study of pronouns shows that all Austronesian languages outside Taiwan belong to a single subgroup, Malayo-Polynesian. These ends he achieves brilliantly, and the findings of this paper support his conclusions (§3.7).

543

Malcolm Ross

Harvey (1982:74–84) provides a critique of earlier reconstructions, and many of his comments are supported by the work reported here. For PAn he reconstructs three sets of pronouns. His nominative set has alternate forms which correspond to my neutral and nominative sets and incorporate the *∅-/*k- alternation. His genitive set also has two variants, one corresponding to my monosyllabic clitics, the other to Blust’s *ni- set (and the comment on these above also applies here). His oblique set corresponds to my accusative set. He also recognises the existence of reflexes of my genitive *m- set but does not reconstruct it, apparently because of inadequacies in the data available to him. One perhaps significant aspect of the data examined in this paper is that, other than the first politeness shift shared by Siraya, Kanakanavu, and PMP, the data appear to display no shared innovations across pairs of Formosan languages, whether previously recognised or not. Blust (1999a) groups Kavalan, Amis and Siraya together, but there is no evidence of shared innovations here (Amis has, like Paiwan, undergone massive restructuring of its pronoun paradigm). Tsuchida (1976) groups Tsou, Kanakanavu and Saaroa together, but again there are no shared pronominal innovations, and the same is true of Saisiyat and Pazeh, subgrouped by Blust. There are a number of innovations, noted at various points in the paper, which are coterminous with Sagart’s (2004) Pituish subgroup. However, to qualify as ‘Pituish’, a feature needs only not to be reflected in Pazeh and Saisiyat, and some such feats may indeed be older but have been lost in these two languages. Two possible innovations, the plural marker *-a and the first politeness shift, introducing forms with *k-, are reflected only in Enemish languages, but not in all of them.

544

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

Appendices

A. Lower-order case-marker reconstructions

This appendix provides supporting data and, in some cases, arguments for lower-order reconstructions of case-markers.

A.1 Proto Atayal (Huang 1993, 1995, 2006, Li 1995, Rau 1992, Tsukida 2005)

Mayrinax Atayal makes a ±SP distinction which I attribute to P-Atayal, as it is needed to account for forms in the other dialects. The latter have lost this distinction (but Wulai and Plngawan keep some contrasting forms) and also the personal case-markers. Li (1995:30) presents Mayrinax/Wulai comparisons, but without reconstructions.

COMMON NOM GEN OBL LOC +SP -SP +SP -SP +SP -SP P-Atayal *kuʔ, *aʔ *na-kuʔ, *naʔ *ci-kuʔ *c(i,u)ʔ, *iʔ *kaʔ *na-kaʔ *ci-kaʔ *caʔ Mayrinax Atayal kuʔ aʔ nkuʔ naʔ ckuʔ cuʔ iʔ Plngawan Atayal kaʔ — nakaʔ naʔ cikaʔ ciʔ — 1 Wulai Atayal quʔ — nquʔ naʔ squʔ sa (suʔ) Teruku Seediq ka — — na — — —

1 Li (1995) has suʔ and saʔ where Rau (1992) and Huang (1993) have only sa.

PERSONAL NOM GEN OBL LOC P-Atayal (*iʔ) (*niʔ) (*iʔ) (*kiʔ), (*∅/*-an) Mayrinax Atayal iʔ niʔ iʔ kiʔ Plngawan Atayal — niʔ — — 1 Teruku Seediq — — — ∅/-an

1 ∅ after a vowel; -an after a consonant.

A.2 Rukai (Li 1997b, Li 1997a, Löbner 1985, Wang 2003, Zeitoun 1995)

The Rukai dialects form three groups, Budai–Labuan–Tanan, Maga–Tona, and Mantauran. The ﬁrst two have noun-phrase markers. All Rukai dialects have accusative alignment, and this has evidently led to shifts in the functions of case-markers.

545

Malcolm Ross

COMMON PERSONAL NOM OBL NOM OBL P-Rukai *ka, *ku, (*na) (*sa), (*-anə), *ki *-anə … PBLT *ka, *ku *ka, *ku, (*sa) *ku *ki Budai ku, ka ka, ku ku, ka ki Labuan ko ko ko ki Tanan ka ka, sa ko ki 1 PMT *na -D?, *ku +D? (*-anə) *ki *-anə Maga na -anɨ ki -a(na), -a(nɨ) 1 Tona na -D, ko +D na, ko ki -anə

1. Tona ko is used with what Löbner (1985) calls pragmatic-definite noun phrases, na with semantic-definites and with indefinites (Wang 2003).

A.3 Proto Puyuma (Stacy Fang-ching Teng, pers. comm., Tsuchida 1980)

Data are available for only two of the perhaps five dialects of Puyuma. Tamalakaw Puyuma has a three-way opposition between definite, indefinite specific and indefinite nonspecific. Nanwang Puyuma has an opposition between definite and indefinite common case-markers, which correspond respectively to the Tamalakaw specific and nonspecific forms. The Tamalakaw definites appear to be innovatory, as they have no known cognates elsewhere. Nanwang has lost GEN forms, the functions of which have been taken over by the obliques. Tamalakaw has just one dedicated common GEN form, [ni]nina GEN:-D:+SP, not shown in the table. This appears to be a recent innovation, a combination of ni GEN:PS and nina GEN/OBL:+D, itself an innovation. The Tamalakaw personal GEN ni, however, is clearly a retention, and also occurs in the personal oblique ka-ni. Nanwang ka-na OBL:PS:P apparently reﬂects an otherwise lost Proto Puyuma *na GEN:PS:P.

COMMON NOM GEN/OBL LOC P-Puyuma — *[i]na +D *a -D — *kana +D *ɖa -D *i Nanwang — [i]na +D a -D — kana +D ɖa -D i 1 Tamalakaw ni +D [i]na -D:+SP a -D:-SP nina +D kana -D:+SP ʐa -D:-SP i

1 Tamalakaw kana OBL:-D:+SP does not appear to be used as GEN.

546

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

PERSONAL SINGULAR PLURAL NOM GEN OBL NOM GEN OBL P-Puyuma *i *ni *ka-ni *na (*na) *ka-na Nanwang i — kan na — kana Tamalakaw i ni kani na ni kani

A.4 Proto Amis (Chen 1987, Tsukida 1993, Wu 1995)

COMMON NEUT NOM GEN ACC LOC P-Amis *u *ku *nu *tu *i […-an] +D? *kiya *iya *[nu]niya *tiya — Sakizaya u ku nu tu i +D kiya iya [nu]niya tiya — Central u ku nu tu i Nataoran u ku nu tu i […-an]

PERSONAL SINGULAR PLURAL NOM GEN ACC LOC NOM GEN ACC P-Amis *ci *ni *ci …-an (*i ci) *ca *na *ca …-an Sakizaya, Central ci ni ci …-an — ca na ca …-an Nataoran1 ci ni — i ci … … …

1 Chen (1987:127) does not list personal plurals.

A.5 Proto Bunun (Jeng 1969, 1977, Wu 1969, Yeh 1999)

The data available for Bunun dialects are in some respects unsatisfactory, as the notes below indicate. Personal case-markers were not found in Jeng’s (1977) Takbanuað data. Yeh (1999) reports that there is no common/personal distinction in Isbukun.

COMMON PERSONAL NOM PSR GEN-OBL NOM PSR, OBL Proto Bunun *a, (ka), (*ca) *itu (*i), (*ki) Takituduh ca =s/=is1 =s/=is1 kat/=at1 =t/=it1 2 2 Takbanuað [k]a, ∅ … [k]i, ∅ … … 3 Isbukun a, ∅ Itu a, mas, a mas, ∅ — —

547

Malcolm Ross

1 The ﬁrst item occurs after a vowel; the second after a consonant. 2 Forms with k- are given in phrase structure rules (Jeng 1977:121) but do not occur in examples. I have found only one example with a NOM (as SBJ with a personal noun in Jeng 1977:88), as both NOM and OBL are usually deleted. Examples with i are more frequent (e.g. Jeng 1977:74, 75, 205). This interpretation agrees with the notes in Jeng (1969:139). Some texts in Jeng (1969:80ff) contain case-markers a and i, but often unglossed. 3 Yeh (1999) says that PSR also marks AGT, but there are no examples of this.

B. Pronominal paradigms of Formosan languages

This appendix provides supporting data for lower-order reconstructions of pronouns and for the reconstructions in the body of the paper.

B.1 Pazeh (Li and Tsuchida 2001:33–37)

1 1 1S 2S 3S 1IP 1EP 2P 3P NEUT y-aku i-siw i-sia i-ta y-ami i-mu y-a-sia 2 NOM aku siw sia ta ami mu a-sia 3 GEN1 — — — ta- — — — GEN2 naki ni-siw ni-sia ni-ta ni-am ni-mu n-a-sia 4 LOC yaku-[n]an i-siw-an i-sia-an i-ta-an y-ami-[n]an i-mu-[n]an — 5 6 Case-marker functions are NEUT=TPC/NSBJ /PAT, NOM=VSBJ, GEN=AGT/PSRN, LOC=LOC.

1 There are three sets of third-person forms, with the roots mini CLOSE, misiw DISTAL and sia INVIS. The first two are not shown here, as they appear irrelevant to PAn reconstruction. 2 From the data, it seems that NOM forms may be postverbal enclitics, but Li and Tsuchida (2001) indicate that there is some freedom in where they occur. 3 The prefix ta- is an AGT form in polite requests like ta-kan-i ‘Let’s eat’. It may co-occur with a long GEN form: ta-kan-i nita ki alaw ‘Let’s eat the fish’ (ki NOM) (Li and Tsuchida 2001: 36–37). See also Ferrell (1968) and Li (1978:582). 4 Locative forms are optionally preceded by di and mean ‘at my place’ etc. 5 Ferrell’s (1968) text has NEUT also as VSBJ; NOM forms occur less often in that text. 6 Adnominally, PSRN precedes the possessum and is followed by a ligature (Li and Tsuchida 2001:35–36).

B.2 Saisiyat (Tunghe) (Yeh 1991:50-53, Huang et al. 1999, Yeh 2003)

All Saisiyat pronouns are free forms. The PSR2 (PSRN) and BEN (BEN, AGT) sets are not shown. The PSR2 set consists of ʔan-ROOT-a[ʔa], where ROOT is as in LOC kan-ROOT.

548

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

The BEN set consists of ʔini-ROOT.

1S 2S 3S 1IP 1EP 2P 3P NOM ya[k]o ʃoʔo sia ʔitaʔ yami moyo la-sia ACC yak-in ʔi-ʃoʔo-n hi-sia, ʔi-sia ʔini-mitaʔ ʔini-yaʔom ʔini-mo-n hi-la-sia PSR1 maʔan ni-ʃoʔ ni-sia m-itaʔ ni-yaʔom ni-mo-n n-a-sia LOC kan-man kan-ʃoʔ kan-sia kan-ʔitaʔ kan-yami kan-moyo kan-la-sia Case-marker functions are NOM=SBJ, ACC=PAT, PSR1= PSRA, LOC= LOC.

B.3 Proto Atayal (Chang 1997, Huang 1989, 1993, 1995, 2006, Pecoraro 1979, Rau 1992, Tsukida 2005)

1S 2S 3S 1IP 1EP 2P 3P 1 2 1 3 3 4 5 5 NEUT: TPC, DISJ, PREDN, SBJ, AGT, PAT, BEN, LOC P-Atayal *(ku(y)iŋ), *isuʔ *hiyaʔ *itaʔ *(cami), *(cimu), *rehiyaʔ, *(i-aku) *(i-ami) *(i-amu) *lahaʔ Mayrinax kuiŋ isuʔ hiyaʔ itaʔ cami cimu nhaʔ Plngawan kuriŋ isuʔ hiyaʔ itaʔ cami cimu lahaʔ Wulai kuziŋ i-suʔ hiyaʔ itaʔ sami simu [l]hɣaʔ Teruku yaku isu hiya ita yami yamu dehiya 6 7 NOM: VSBJ P-Atayal *=[ca]ku *= suʔ — *= taʔ *(=cami) *(=cimu) — Mayrinax =cu =suʔ — =taʔ =cami =cimu — 8 8 Plngawan =cu =suʔ — =taʔ =min =mamu Wulai =[sa]kuʔ =suʔ — =taʔ =sami =simu — Teruku =ku =su — =ta =nami,8 =namu8 — =mian 6 GEN: AGT, PSRA P-Atayal *=ku, *=suʔ *=nia, *=taʔ *=n-ami, *(=m-amu), *(=na-haʔ) *=[m-a]ku *=na *=ni-am *(=n-amu) Mayrinax =mu =suʔ — =taʔ =niam =mamu =nhaʔ 9 Plngawan =mu =suʔ =niʔ =taʔ =min =mamu =nahaʔ 9 Wulai =[ma]kuʔ =suʔ — =taʔ =mian =mamu =nhaʔ =mu Teruku =mu =su — =ta =nami, =namu =deha =mian 9 PSR: PSRN Teruku [ne]naku [ne]nisu ne-hiya [ne]nita [ne]nami [ne]namu ne-dehiya

549

Malcolm Ross

OBL:PAT, LOC, PSRN P-Atayal *ke-nan *su-nan *hiya-an *ita-[na]n *(ca)mi-nan *(ca)mu-nan *(re)hiya-an Plngawan ki-nan si-nan hiya-n ita-n cami-nan cimu-nan laha-n Wulai k-nan su-nan hiya-n ita-n smi-nan smu-nan hɣa-n Teruku ke-nan su-nan hiya-ante-nan menani mu-nan dehiya-an

1 Preceded by iʔ in Mayrinax, but not preceded by a case-marker in Wulai (Rau 1992:141). 2 In Mayrinax. 3 Preceded by a NOM case-marker. 4 Preceded by Mayrinax iʔ NOM. Not preceded by a case-marker in Teruku. 5 In Mayrinax, BEN preceded by niʔ, LOC by kiʔ. 6 Both sets are optional agreement pronouns in Teruku, but evidently not in Mayrinax or Wulai. 7 Mayrinax SBJ. 8 Copied from the GEN forms. 9 According to Pecoraro (1979:65) and Chang (1997:13-16) this set is used as PSRN. I assume the addition of ne- (cognate with Mayrinax niʔ—see note 4) to be of post-P-Atayal antiquity, as the third-person forms to which it is attached are NEUT forms.

B.4 Thao (Blust 2003)

1S 2S 3S 1IP 1EP 2P 3P 1 NOM1 yaku ihu, [y]uhu θiθu ita yamin maniun θayθuy NOM2 wak — — — — — — 2 GEN1 nak m-ihu θiθu m-ita yamin, nam maniun θayθuy 3 GEN2 -[a]k, -[i]k — — — — — — 2 2 ACC yakin, ihu-n, uhu-n , θiθu-n ita-n, yamin, ami-n, maniun θayθuy nakin4 m-ihu-n3 uta-n2 namin4 5 6 Case-marker functions are NOM1=VSBJ, GEN1=AGT/PSRN/PSRA, ACC=PAT; NOM2 wak 1S serves as PAT of imperative and VSBJ of a future actor-voice verb.

1 Although there is a clear contrast between NOM and ACC forms, NOM forms are sometimes used for PAT where ACC is expected (Li 1978:598, Blust 2003:207). 2 Form from Li (1978). 3 -[i]k may reﬂect earlier -in PV+GEN:1S, and -[a]k earlier -an LV+GEN:1S (Li 1978:599, Blust 2003:92-93, 96-97). However, the contrast is blurred in modern Thao, and these affixes also occur occasionally as AVSBJ. 4 This form is a rare variant which sometimes replaces the first item shown when it occurs as a benefactive. Blust (2003:208) comments that m-ihu-n 2S is a combination of GEN and OBL forms.

550

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

5 Usually a GEN possessor pronoun precedes its possessum noun with an optional ligature, i.e. it behaves as PSRN. Occasionally it follows (= PSRA). 6 I have not found any indication whether these forms also have LOC function.

B.5 Siraya (Adelaar 1997)

The orthography is Adelaar’s: ĭ represents a short i.

1S 2S 3S 1IP 1EP 2P 3P NEUT ĭau ĭmhu təni ĭmĭtta ĭmian ĭmumi neini NOM =koh =kow — =kĭtta =kame =kamu — 1 GEN =[m]au =[m]uhu tĭn =[m]ĭtta =[m]ian =[m]umi nein OBL ĭau-an ĭmhu-an tĭni-an ĭmĭtta-n ĭmian-an ĭmumi-an neini-an Case-marker functions are NEUT=SBJ/TPC, NOM=VSBJ?, GEN=AGT, PSRA, ACC=PAT.

1 m- after a vowel.

B.6 Tsou (Zeitoun 2005)

There are no third-person personal pronouns, as they are replaced by demonstratives. 1S 2S 1IP 1EP 2P NEUT aʔo suu aʔati aʔami muu 1 2 NOM-GEN -ʔo/-ʔu -su/-ko -to -mza -mu 3 4 Case-marker functions are NEUT=VSBJ/PAT , NOM- GEN=AVSBJ, AGT, PSRA.

1 -ʔo follows -i; -ʔu occurs elsewhere. 2 The difference between -su and -ko is largely dialectal. 3 As subject a free pronoun is optionally preceded by the NOM marker na. As PAT this never occurs. (Tung 1964 and Tsuchida 1976:98 show these forms with na- affixed in all contexts.) 4 AGT pronouns are agreement markers.

551

Malcolm Ross

B.7 Kanakanavu (Li 1997a, Mei 1982, Tsuchida 1976:38)

1S 2S 3S/P 1IP 1EP 2P NEUT ii-ku, ii-kasu, — ii-kita ii-kimi, ii-kamu, ii-kia1 ii-mu-kasu1 ii-kia1 ii-mu-kamu1 3 NOM =ku, =kasu =∅, =ini =kita =kimi, =kamu =kia1 =kia1 3 2 GEN1 =[m]aku =[mu]su =kiai, =ini =[mi]ta =mia =mu 4 GEN2 =naku =nsu =nni =nta =nmia — 5 OBL ʔiku-an kasu-an ʔini-an kita-nan kimi-an kamu-an 6 7 8 Case-marker functions are DISJ, AVSBJ, NOM:VSBJ, GEN1/GEN2:AGT, PSRA, OBL: PAT, BEN.

1 The forms ii-kia and =kia are contrastive; ii-mu-kasu and ii-mu-kamu are vocative. 2 =kiai with an IMPF verb, =ini elsewhere (Mei 1982:219). Mei says there is no third-person PSR form, but cina=ini ‘his mother’ in an example on p211 suggests that there is. 3 Generally -m(V) after a vowel, non-m(V) forms after a consonant, but there are unexplained exceptions (Mei 1982:210). 4 Li (1997b:354) records n-initial GEN2 forms that are missing from other sources. He does not know of a functional distinction betwen GEN1 and GEN2. 5 Final -n is deleted except before a following vowel-initial clitic. 6 In marked sentence-initial position, evidently as a topic or focus-fronted, to judge from Mei’s examples. It seems likely that NEUT forms have other functions not illustrated in the data. 7 No example of a verbless clause witha personal pronoun subject has been found in the data. 8 There are three voices: actor, patient, and location. There is a fourth voice, labelled ‘special’ by Tsuchida (1976:49–51) and ‘object focus 2’ by Mei (1982), with unusual argument marking: the agent is encoded as OBL (there is no example with a personal pronoun subject).

B.8 Saaroa (Tsuchida 1976:68)

1S 2S 3S/P 1IP 1EP 2P NEUT iɬa-ku iɬa-u iɬa-isa iɬa-ta iɬa-ɬamu iɬa-mu NOM =aku =u ∅ =ita =amu =mu 1 ACC — — isa=na — — — GEN =ku =u =isa =ta =ɬamu =mu Case-marker functions are NEUT=FOCFR/TPC/DISJ/PAT, NOM=SBJ, ACC=PAT, GEN=AGT, PSRA.

1 There is a dedicated ACC (PAT) form only in 3S. Elsewhere na OBL + NEUT is used.

552

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

B.9 Northern Paiwan (Huang et al. 1999, Anna Hsiou-chuan Chang, pers. comm.)

Data from other dialects are almost identical with Northern Paiwan.

1S 2S 3S 1IP 1EP 2P 3P NEUT ti-aken ti-sun ti-maju ti-cen ti-amen ti-mun ti-a-maju NOM =[a]ken =[e]sun — =[i]cen =[a]men =[e]mun — GEN ku= su= — ca= nia= nu= — PSR ni-aken ni-sun ni-maju n-icen ni-amen ni-mun ni-a-maju OBL canu-aken canu-sun cai-maju canu-icen canu-amen canu-mun cai-a-maju 1 Case-marker functions are NEUT=FOCFR/TPC/DISJ, NOM=SBJ, GEN=AGT/PSRA, PSR= PSRN/AGT, OBL=PAT/ LOC.

1 Third-person members of this set also occur in AGT function.

B.10 Proto Rukai (Zeitoun 1997, Zeitoun 2003, Wang 2003)

Rukai third-person pronouns are deictics and are omitted here. BLT forms are those found in the Budai, Labuan and Tanan dialects. In Labuan and Tanan /u/ is orthographic o. Oblique forms in Budai end in -anə, in Labuan and Tanan in -a. Otherwise there are only minor differences among these dialects.

1S 2S 1IP 1EP 2P NEUT: TPC P-Rukai *aku, *naku *musu *mita *na[m]i *mumu, *nimu BLT ku-naku ku-su ku-ta ku-nai ku-numi Mantauran i-ɭaə i-miaʔə i-mitə, i-ta i-namə i-nomə P-MT *k-akə *musu, *ku-su *mita, *ki-ta *ki-namə *mumu, *ku-mu Maga kɨkɨ musu miti knamɨ mumu Tona kakə koso kita kinamə komo NOM: SBJ P-Rukai *naku, *ku *su, *=musu *ta, *=mita *na[m]i *nimu, *mu= BLT naw-, -[n]aku =su =ta =nai =numi Mantauran -ɭao -moʔo -[mi]ta -nai -nomi P-MT *ku=, *-kə *su= *ta= *namə=, *=namə *mu= Maga ku=, =kɨ so= ta= namɨ=, =namɨ mu= Tona ko= su= ta= namə= mo=

553

Malcolm Ross

PSR: PSRA P-Rukai *=li *=su *=ta *=na[m]i *nimu, *=mu BLT =li =su =ta =nai =numi Mantauran -li -ʔo -ta -nai -nomi P-MT *-li *-su *-ta *-namə *-mu Maga =li =su =ta =namɨ =mu Tona =[i]li =[i]so =[i]ta =[i]namə =[i]mo OBL: PAT, peripheral roles P-Rukai *naku-a[nə] *musu-a[nə], *su-a[nə] *mita-a[nə] *na[m]i-a[nə] *[ni]mu-a[nə] BLT naku-a[nə] musu-a[nə] mita-a[nə] nai-a[nə] numi-a[nə] Mantauran i-a-ə i-miaʔ-ə i-mit-ə i-nam-ə i-nom-ə P-MT *naku-a *su-a, *musu-a *miti-a *nami-a *mu-a[nə] Maga ŋku-a su-a mti-a nma-a mu-a Tona [na]ko-a moso-a miti-a nami-a mo-anə

B.11 Proto Puyuma (Stacy Fang-ching Teng, pers. Comm., Tsuchida 1980)

1S 2S 3S 1IP 1EP 2P 3P NEUT: TPC, FRFOC, DISJ P-Puyuma *i-(ŋ)ku *i-[n]u *(i-ʐiw), *i-(n)ta *i-niam *i-(n)mu *naɖiw *(i-taw) Tamalakaw i-ŋku i-nu i-ʐiw i-nta i-niam i-nmu naʐiw 1 Nanwang ku-i-ku yu-yu ta-y-taw ta-i-ta mi-mi mu-i-mu ∅ NOM:SBJ P-Puyuma *=ku *=(y)u *∅ *=ta *=mi *=mu *∅ Tamalakaw =ku =u ∅ =ta =mi =mu ∅ Nanwang =ku =yu ∅ =ta =mi =mu ∅ 2 GEN:AGT P-Puyuma *ku= *nu= *taw= *ta= *mi= *mu= *taw= Tamalakaw ku= nu= taw= ta= mi= mu= taw= Nanwang ku= nu= tu= ta= mi= mu= tu= PSR1: PSRA P-Puyuma *=li *=u *=taw *(=ta)*(=mi) *(=mu)*(ni-naɖiw) 3 Tamalakaw =li =u =taw =ta =mi =mu ni-naʐiw Nanwang =li =u =taw — — — —

554

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

4 PSR2: PSRA P-Puyuma *ku= *nu= *taw= *ta= *niam= *mu= *taw= Tamalakaw =[ŋ]ku =[ŋ]nu =ntaw =nta =niam =nmu =ntaw Nanwang ku nu tu ta niam mu tu GEN-OBL: AGT, PAT P-Puyuma *kan-(iŋ)ku *kan-nu *kan-taw *kan-ta *ka-niam *kan-mu *kan-(an-)taw Tamalakaw kan-iŋku kan-nu kan-taw kan-ta ka-niam kan-mu kan-an-taw Nanwang kan-ku ka-nu kan-taw kan-ta ka-niam kan-emu kan-taw

1 Nanwang NEUT forms seem to represent a restructuring with reduplication of the base in front of the *i- prefix, e.g. ku-i-ku NEUT:1S< *ku+*i-(ŋ)ku. 2 GEN acts as an agreement marker in both dialects, as it remains even when there is a GEN (Tamalakaw) or OBL (Nanwang) NP actor or possessor. 3 In Nanwang these forms occur with only a very small number of kin terms, and only in the singular. It is possible that Tamalakaw non-singulars are simply copied from NOM (3S from NEUT) and did not occur in Proto Puyuma. 4 In Tamalakaw the forms in the PSR2 set are attached to a case-marker and precede the possessumnoun. In Nanwang, they either (i) are proclitic to the possessum, forming a NOM NP, or they are enclitic to the case-marker which precedes the possessum. I infer that in Proto Puyuma these forms were proclitic to the possessum, but optionally preceded by a case-marker.

B.12 Amis (Chen 1987:135–136, Fey 1986:381, Tsukida 1993, Wu 1995)

Forms and functions are so similar across dialects that differences are simply noted below the table.

1 1 1S 2S 3S 1IP 1EP 2P 3P 2 NEUT aku isu c-ira [k]ita ami amu uhni NOM k-aku k-isu c-ira k-ita k-ami k-amu k-uhni 3 GEN1 =aku =isu — =ita — =amu =uhni 4 GEN2 m-aku, m-isu, n-ira m-ita niam n-amu n-uhni (n-aku) (n-isu) 5 OBL [i]t-aku-an [i]t-isu-an [i]c-ira-an [i]t-ita-an, [i]t-aman [i]t-amu-an [i]t-uhnan [i]kita-an Case-marker functions are NEUT=PREDN/TPC, NOM=SBJ, GEN1= PSRA/AGT, GEN2= 5 AGT/PSRA/PSRN, OBL=PAT/LOC.

1 Third-person forms vary from dialect to dialect and are related to demonstratives.

555

Malcolm Ross

2 In Sakizaya a (clause-initial) predicate NP (PREDN) is formed with u (C:NEUT) + PRON:NEUT, i.e. u-aku etc. The NEUT set does not occur in the Central data, where the NOM set takes over its functions. 3 Only Chen (1987:136) analyses this set as enclitics. As they were almost certainly enclitics in PAn, they were probably enclitics in Proto Amis. They function as PSRA in all three dialects, and also as AGT in Central Amis. I infer that they had AGT function in Proto Amis, inherited from PAn. 4 GEN2 forms (exept for n-initial 1S and 2S) are usually preceded by nu C:GEN. The n-initial 1S and 2S forms are only in Nataoran (as optional variants), and are not reconstructable with certainty to Proto Amis. 5 In the limited data in Tsukida (1993) the Sakizaya OBL forms have only PAT function, but in Nataoran and Central they are also LOC. In Sakizaya and Nataoran i- apparently obligatorily precedes them; in Central it is optional. In Sakizaya and Central, -an often becomes -anan.

B.13 Proto Bunun (Jeng 1969, 1977, Wu 1969, Yeh 1999)

Third-person forms are not reconstructed as these vary considerably from dialect to dialect and appear all to be demonstratives.

1S 2S 1IP 1EP 2P 1 NEUT: FRFOC, TPC, DISJ, PAT?, AGT P-Bunun *ðaku *suʔu *ita *ðami *mu[ʔu] Takituduh ðako soʔo ita ðami moʔo Takbanuað ðaku suʔu ʔitaʔ ðamiʔ muʔu Isbukun ðaku su mita ðami mu NOM1: SBJ P-Bunun *ka-ðak, (*saikin) *ka-su *ka-ta *ka-ðam *ka-mu Takituduh a-ðak a-so a-ta a-ðam a-mo Takbanuað ʔaðak ʔa-suʔ ʔa-taʔ ʔa-ðam ʔa-muʔ Isbukun saikin ka-su ka-ta ka-imin ka-mu NOM2: VSBJ P-Bunun *=[s]a-k *=a-s *=[a-]ta *=[s]a-m *=a-mu Takituduh =[k]a-k =[k]a-s =[k]a-ta =[k]a-m =[k]a-mo Takbanuað =sak =suʔ =taʔ =sam — Isbukun =ik =as =ta =im =am 2 ACC: PAT P-Bunun *ðak-un *suʔu-n *it-un *ðam-un *muʔu-n 3 Takituduh ðaku-n soʔu-n, a-so itu-n, ita ðamu-n, ðami moʔu-n

556

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

Isbukun ma[s]-ðaku ma-su ma[s-m]ita ma[s]-ðami ma[s]-mu 4 GEN: AGT, PAT P-Bunun *=ku *=su[ʔu] *=ta *=ðami *=mu[ʔu] Takituduh =ko =soʔo =ta =ðami =moʔo Takbanuað =ku =su =taʔ — =mu Isbukun =ku =su =ta — =mu 5 PSR: PSRA, PSRN P-Bunun *=nak *=suʔu *=nita, *=mita *=nam *=nuʔu, *=mu[ʔu] 3 Takituduh =nak =soʔo =nita, =mita =nam =noʔo, =mu Takbanuað nak suʔu =mita =nam =muʔu Isbukun =nak =su =mita =nam =mu 6 LOC: LOC P-Bunun *ðaku-an *suʔu-an *ita-an *ðami-an *muʔu-an Takituduh ðaku-an suʔu-an ita-an ðami-an muʔu-an Isbukun ʔi=ðaku-an ʔi=su-an ʔi=mita-n ʔi=ðami-an ʔi=mu-an

1 In Takbanuað a NEUT form may apparently be preceded by the OBL marker [k]i if it serves as AGT or PAT (one AGT example, Jeng 1977:206). 2 In Takituduh these forms are ACC, in Isbukun GEN/ACC. There is no corresponding set in Takbanuað. 3 Where a second Takituduh form is shown, it is Li’s. 4 In Takituduh GEN enclitics appear to be agreement markers. 5 In Takituduh PSR forms are either PSRA enclitics or are prefixed with [k]i- to form PSRN. In Takbanuað they appear to be PSRA enclitics (Jeng is not explicit about cliticisation), or are prefixed with i- to form PSRN. In Isbukun (where they have largely merged with GEN) they are prefixed with ʔi- to form PSRN. 6 There is no corresponding set in Takbanuað.

557

Malcolm Ross

B.14 Kavalan (Chang 1997:32-34, Chang 2000:84-99, Lee 1997, Li and Tsuchida 2006, Tsuchida 1993)

1S 2S 3S1 1IP 1EP 2P 3P 2 NEUT a-iku a-isu ti-yau a-ita a-imi a-imu qan-iau NOM =iku =su — =ita =imi =imu — ACC ti-ma-iku ti-ma-isu ti-yau ti-ma-ita ti-ma-imi ti-ma-imu qan-iau 3 GEN -ku -su -na -ta -niaq -numi -na PSR za-ku za-su za-na za-ta za-imi, za-imu, za-na 4 4 za-niqa za-numi 5 LOC ta-ma-iku-an ta-ma-isu-an ti-yau-an ta-ma-ita-an ta-ma-imi-an ta-ma-imu-an ta-qan-iau-an 6 Case-marker functions are NEUT=SBJ/TPC/FRFOC, NOM=VSBJ, ACC=PAT, GEN=AGT, PSR= PSRA/PSRN, LOC= LOC.

1 Alternant 3S free forms have the root izip ‘body’ (Tsuchida 1993): a-izipna NEUT, ti-ma-izipana ACC, ta-ma-izip-an LOC. 2 qani plural marker, iau ‘that’ (Tsuchida 1993). 3 Optional agreement suffixes (may not occur with non-human referent) (Chang 1997:119-120). On the suffix/enclitic contrast, see Lee (1997:38-44). 4 Forms from Li and Tsuchida (2006). 5 Li and Tsuchida (2006) have ti- in ACC where Chang has ta- and ta- in LOC where Chang has ti-. 6 FRFOC example in Lee (1997:48).

558

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

References

Adelaar, K. Alexander. 1997. Grammar notes on Siraya, an extinct Formosan language. Oceanic Linguistics 36:362-397. Billings, Loren, and Daniel Kaufman. 2004. Towards a typology of Austronesian pronominal clisis. Proceedings of the Eleventh Meeting of the Austronesian Formal Linguistics Association, ed. by Paul Law. ZAS Papers in Linguistics 34. Berlin: Zentrum für Allgemeine Sprachwissenschaft. Blust, Robert A. 1977. The Proto-Austronesian pronouns and Austronesian subgrouping: a preliminary report. University of Hawai‘i Working Papers in Linguistics 9.2:1-15. Blust, Robert A. 1995. Austronesian comparative dictionary. Computer files. Honolulu: University of Hawai‘i. Blust, Robert A. 1999a. Notes on Pazeh phonology and morphology. Oceanic Linguistics 38:322-365. Blust, Robert A. 1999b. Subgrouping, circularity and extinction: some issues in Austro- nesian comparative linguis-tics. Selected Papers from the Eighth International Conference on Austronesian Linguistics, ed. by Elizabeth Zeitoun and Paul Jen-kuei Li, 31-94. Taipei: Institute of Linguistics, Academia Sinica. Blust, Robert A. 2003. Thao Dictionary. Taipei: Institute of Linguistics, Academia Sinica. Blust, Robert A. 2005. A note on the history of genitive marking in Austronesian languages. Oceanic Linguistics 44:215-222. Chang, Yung-li. 1997. Voice, Case and Agreement in Seediq and Kavalan. Hsinchu: National Tsing Hua University dissertation. Chang, Yung-li. 2000. Kavalan Reference Grammar. Taipei: Yuanliu. (in Chinese) Chang, Yung-li. 2004. AF verbs: transitive, intransitive, or both? Studies in Sino-Tibetan Languages: Papers in Honor of Professor Hwang-cherng Gong on His Seventieth Birthday, ed. by Ying-chin Lin, Fang-min Hsu, Chin-chih Lee, Jackson T.-S. Sun, Hsiu-fang Yang and Dah-an Ho, 95-119. Taipei: Institute of Linguistics, Academia Sinica. Chen, Teresa M. 1987.Verbal Constructions and Verbal Classification in Nataoran-Amis. Pacific Linguistics C-85. Canberra: The Australian National University. Dahl, Otto Christian. 1973. Proto-Austronesian. Oslo: Studentlitteratur. Ferrell, Raleigh. 1968. The Pazeh-Kahabu language. Bulletin of the Department of Archaeology and Anthropology, 31/32:73-96. Taipei: National Taiwan University. Fey, Virginia. 1986. Amis Dictionary. Taipei: The Bible Society in the Republic of China. Harvey, Mark. 1982. Subgroups in Austronesian. Papers from the Third International

559

Malcolm Ross

Conference on Austronesian Linguistics 2, Tracking the Travellers, ed. by Amran Halim, Lois Carrington and S.A. Wurm, 47-99. Pacific Linguistics C-75. Canberra: The Australian National University. Himmelmann, Nikolaus P. 2002. Voice in western Austronesian: an update. The History and Typology of Western Austronesian Voice Systems, ed. by Fay Wouk and Malcolm Ross, 7-16. Pacific Linguistics 518. Canberra: The Australian National University. Himmelmann, Nikolaus P. 2005. The Austronesian languages of Asia and Madagascar: typological characteristics. The Austronesian Languages of Asia and Madagascar, ed. by Alexander Adelaar and Nikolaus P. Himmelmann, 110-181. London: Routledge. Huang, Lillian, Elizabeth Zeitoun, Marie M. Yeh, Anna H. Chang and Joy J. Wu. 1999. A typological overview of pronominal systems of some Formosan languages. Proceedings of the Fifth International Conference on Chinese Linguistics, ed. by Hsu Wang, Feng-fu Tsau and Chin-fa Lien, 165-198. Taipei: Crane. Huang, Lillian M. 1989. The pronominal system of Atayal. Studies in English Literature and Linguistics May 1989:115-133. Huang, Lillian M. 1993. A Study of Atayal Syntax. Taipei: Crane. Huang, Lillian M. 1995. A Study of Mayrinax Atayal. Taipei: Crane. Huang, Lillian M. 2006. Manifestations of participants in Atayal: a cross-dialectal Study. Paper presented at the Tenth International Conference on Austronesian Linguistics. Puerto Princesa, Philippines. Jeng, Heng-hsiung. 1969. A preliminary report on a Bunun dialect as spoken in Hsinyi, Nant’ou, Taiwan. Mimeo, National Taiwan University, Taipei. Jeng, Heng-hsiung. 1977. Topic and Focus in Bunun. Taipei: Institute of History and Philology, Academia Sinica. Lee, Amy Peijung. 1997. The Case-marking and Focus Systems in Kavalan. Hsinchu: National Tsing Hua University thesis. Li, Paul Jen-kuei. 1973. Rukai Structure. Taipei: Institute of History and Philology, Academia Sinica. Li, Paul Jen-kuei. 1978. The case-marking systems of the four less-known Formosan languages. Second International Conference on Austronesian Linguistics: Proceedings, ed. by S.A. Wurm and Lois Carrington, 569-615. Pacific Linguistics C-61. Canberra: The Australian National University. Li, Paul Jen-kuei. 1995. The case-marking system in Mayrinax, Atayal. Bulletin of the Institute of History and Philology 66:23-52. Li, Paul Jen-kuei. 1997a. A syntactic typology of Formosan languages: case markers on nouns and pronouns. Chinese Languages and Linguistics: Typological Studies of

560

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

Languages in China, ed. by Chiu-yu Tseng, 343-378. Taipei: Institute of History and Philology, Academia Sinica. Li, Paul Jen-kuei. 1997b. The Tona dialect. The Formosan Languages of Kaohsiung, ed. by Paul Jen-kuei Li, 119-158. Kaohsiung: Kaohsiung County Government. (in Chinese) Li, Paul Jen-kuei, and Shigeru Tsuchida. 2001. Pazih Dictionary. Taipei: Institute of Linguistics, Academia Sinica. Li, Paul Jen-kuei, and Shigeru Tsuchida. 2006. Kavalan Dictionary. Taipei: Institute of Linguistics, Academia Sinica. Liao, Hsiu-chuan. 2004. Transitivity and Ergativity in Formosan and Philippine Languages. Honolulu: University of Hawai‘i dissertation. Löbner, Sebastian. 1985. Definites. Journal of Semantics 4:279-326. Mei, Kuang. 1982. Pronouns and verb inﬂection in Kakakanavu. Tsing Hua Journal of Chinese Studies, New Series 14:207-231. Pecoraro, Ferdinando. 1979. Eléments de grammaire taroko, précédés del la présentation de la vie et de la culture des taroko. Paris: Association Archipel. Radetzky, Paula. 2003. Grammaticalization of a definiteness marker in Saaroa. University of Oregon. Manuscript. Rau, Der-hwa Victoria, 1992. A grammar of Atayal. Taipei: Crane. Reid, Lawrence A. 1978. Problems in the reconstruction of Proto-Philippine construction markers. Second International Conference on Austronesian Linguistics: Proceedings, Fascicle 1, ed. by S.A. Wurm and Lois Carrington, 33-66. Pacific Linguistics C-61. Canberra: The Australian National University. Reid, Lawrence A. 1982. The demise of Proto-Philippines. Papers from the Third International Conference on Austronesian Linguistics 2, Tracking the Travellers, ed. by Amran Halim, Lois Carrington and S.A. Wurm, 201-216. Pacific Linguistics C-75. Canberra: The Australian National University. Reid, Lawrence A., and Hsiu-chuan Liao. 2004. A brief syntactic typology of Philippine languages. Language and Linguistics 5.2:433-490. Ross, Malcolm. 2002a. The history and transitivity of western Austronesian voice and voice-marking. The History and Typology of Western Austronesian Voice Systems, ed. by Fay Wouk and Malcolm Ross, 17-62. Pacific Linguistics 518. Canberra: The Australian National University. Ross, Malcolm. 2002b. Takia. The Oceanic Languages, ed. by John Lynch, Malcolm Ross and Terry Crowley, 216-248. Richmond: Curzon Press. Ross, Malcolm. 2005. The Batanic languages in relation to the early history of the Malayo-Polynesian subgroup of Austronesian. Journal of Austronesian Studies 1/2:1-24.

561

Malcolm Ross

Ross, Malcolm, and Stacy Fang-ching Teng. 2005. Formosan languages and linguistic typology. Language and Linguistics 6.4:739-781. Sagart, Laurent. 2004. The higher phylogeny of Austronesian and the position of Tai-Kadai. Oceanic Linguistics 43:411-444. Starosta, Stanley. 1986. Focus as recentralisation. FOCAL I: Papers from the Fourth International Conference on Austronesian Linguistics, ed. by Paul Geraghty, Lois Carrington and S.A. Wurm, 73-95. Pacific Linguistics C-93. Canberra: The Australian National University. Starosta, Stanley. 1992. The case-marking system of Proto-Formosan. Paper presented at the Third International Symposium on Language and Linguistics: Pan-Asiatic Linguistics. Bangkok. Starosta, Stanley, Andrew K. Pawley, and Lawrence A. Reid. 1981. The evolution of focus in Austronesian (full paper). Paper presented at the Third International Conference on Austronesian Linguistics. Bali. Starosta, Stanley, Andrew K. Pawley, and Lawrence A. Reid. 1982. The evolution of focus in Austronesian. Papers from the Third International Conference on Austronesian Linguistics 2, Tracking the Travellers, ed. by Amran Halim, Lois Carrington and S.A. Wurm, 145-170. Pacific Linguistics C-75. Canberra: The Australian National University. Tsuchida, Shigeru. 1976. Reconstruction of Proto-Tsouic Phonology. Studies of Languages and Cultures of Asia and Africa, Monograph Series 5. Tokyo: Institute for the Study of Languages and Cultures of Asia and Africa. Tsuchida, Shigeru. 1980. Puyuma-go (Tamarakaohōgen) goi: fu gohōgaisetsu oyobi tekisuto [Puyuma (Tamalakaw dialect) vocabulary: with grammatical notes and texts]. Kuroshio no minzoku, bunka, gengo [Ethnology, Cultures and Languages along the Black Current], ed. by Kuroshio Bunka no Kai [Black Current Cultures Committee], 183-307. Tokyo: Kadokawa Shoten. Tsuchida, Shigeru. 1993. Short sketch of Kavalan grammar. Manuscript. Tsukida, Naomi. 1993.A brief sketch of the Sakizaya dialect of Amis. Tokyo University Linguistic Papers 13:375-413. Tsukida, Naomi. 2005. Seediq. The Austronesian Languages of Asia and Madagascar, ed. by Alexander Adelaar and Nikolaus P. Himmelmann, 291-325. London: Routledge. Tung, T’ung-ho. 1964. A Descriptive Study of the Tsou Language. Taipei: Institute of History and Philology, Academia Sinica. Wang, May Hsiu-mei. 2003. Morpho-syntactic Manifestations of Participants in Tona (Rukai). Taipei: National Taiwan University MA thesis. Wolff, JohnU. 1996. The development of the passive verb with pronominal prefix in western Austronesian languages. Reconstruction, Classification, Description:

562

Reconstructing the Case-marking and Personal Pronoun Systems of Proto Austronesian

Festschrift in Honor of Isidore Dyen, ed. by Bernd Nothofer, 15-40. Hamburg: Abera. Wu, Joy. 1995. Referential choice in Amis narrative: a case study. Research Papers in Linguistics and Literature 4:211-230. Wu, Peter. 1969. A descriptive analysis of Bunun language. Georgetown University, Washington DC. Manuscript. Yeh, Marie Meili. 1991. Saisyat Structure. Hsinchu: National Tsing Hua University MA thesis. Yeh, Marie Meili. 1999. Bunun reference grammar. Manuscript. Yeh, Marie Meili. 2003. A Syntactic and Semantic Study of Saisiyat Verbs. Taipei: National Taiwan Normal University dissertation. Zeitoun, Elizabeth. 1995. Problèmes de linguistique dans les langues aborigènes de Taiwan. Université de Paris 7 dissertation. Zeitoun, Elizabeth. 1997. The pronominal system of Mantauran (Rukai). Oceanic Linguistics 36:312-346. Zeitoun, Elizabeth. 2003. Toward a reconstruction of Proto-Rukai morpho-syntax. Paper presented at the Tenth Conference of the Austronesian Formal Linguistics Association. Honolulu: University of Hawai‘i. Zeitoun, Elizabeth. 2005. Tsou. The Austronesian Languages of Asia and Madagascar, ed. by Alexander Adelaar and Nikolaus P. Himmelmann, 259-290. London: Routledge. Zeitoun, Elizabeth. 2006. On the notion of plurality in Formosan languages and its incidence in PAn reconstruction. Taipei: Institute of Linguistics, Academia Sinica. Manuscript.

563