The position of Chamorro and Palauan in the Austronesian family tree: evidence from verb morphosyntax ERIK ZOBEL 1 Introduction I Palauan and Chamorro, spoken in Micronesia on the Palau (Belau) and Mariana Islands, resectively, have long een recognised as outliers in the Pacific region, with stronger ties to the languages of the Philippines and Indonesia than to neighbouring languages. In his ground-breaking monograph, Dempwolff ( 1 934-3 8) divided the Austronesian language family into an 'Indonesian' and a 'Melanesian' subgroup. It has ecome apparent that according to this cheme, Palauan and Chamorro must e included in the 'Indonesian' subgroup, since they do not share the innovations characterising the 'Melanesian' subgroup (which latter under the lael 'Oceanic' has remained firmly established as a well-defined subgroup of the Austronesian family). In Dyen's lexicostatistical classification of the Austronesian languages ( 1 965), Chamorro and Palauan are isolates of the 'Malayo-Polynesian' linkage, coordinate to subgroups of relatively high order. In spite of the problematic nature of lexicostatistics (Dyen's classification fails to recognise well-established subgroups such as Oceanic), it illustrates the isolated character of Palauan and Chamorro with regard to their lexicon. Blust ( 1 977) proosed a classification of the Austronesian languages which up to now has gained wide acceptance (Figure 1 ). It contains two ndes relevant to the discussion here: the Malayo-Polynesian (MP) subgroup, based on phonological, lexical and grammatical innovations (Blust 1 995); and the Central-Easten Malayo-Polynesian (CEMP) subgroup (chiefly based on lexical innovations; Blust 1 993), which includes the Central M a layo­ Polynesian (CMP) and the Easten Malayo-Polynesian (EMP) group. The latter contains the I am indebted to Bernd Nothofer, Malcolm Ross and Sander Adelaar for their comments on earlier drafts of this paper. Fay Wouk and Mlcolm Rss, s, e loy ad poloy of wtn Atonan e ts, 405-434. Canera: Pcic Linguistis, 202. © Erik el Zobel, E. "The position of Chamorro and Palauan in the Austronesian family tree: evidence from verb morphosyntax". In Wouk, F. and Ross, M. editors, The history and typology of western Austronesian voice systems. PL-518:405-436. Pacific Linguistics, The Australian National University, 2001. DOI:10.15144/PL-518.405 ©2001 Pacific Linguistics and/or the author(s). Online edition licensed 2015 CC BY-SA 4.0, with permission of PL. A sealang.net/CRCL initiative. 405 406 Erik Zobel Oceanic languages. MP languages that are not included in CEMP were groued together by Blust in a Westen MP (WMP) group. Austronesian � � various Formosan subgroups Malayo-Polynesian Westen Malayo-Polynesian (incl. Palauan and Chamorro) Central-Easten Malayo-Polynesian Figure 1 : Austronesian family tree (following Blust) According to this classification, Chamorro and Palauan must be included in the WMP subgroup, since they are clearly MP languages, and do not share the innovations defining CEMP. The next task is to establish closer ties etwen thee languages and other WMP languages. The phonological histories of the two languages give no clue apart from showing that they are non-Oceanic Malaya-Polynesian languages. The sound changes of oth are either found n many other WMP languages (Chamorro: merger of *e and *u, Palauan: merger of *D and *Z, *n and *n) or are unique (Chamoro: merger of *D and *k, *j and *q; Palauan: merger of *j and *R, vocalisation of *1 and *p). There are no phonological innovations common to both languages, apart from trivial ones (loss of *S, stress on the PMP enultimate). Unlike Chamoro, Palauan shares with neighbouring Yaese and many Nuclear Micronesian languages the loss of final vowels, which is an areal feature in that pat of the Pacific. There have also een attempts to ue grammatical evidence to establish the closer affiliations of Palauan and Chamorro. Patzold ( 1 968) demonstrated that many Palau an affixes (verbal and nominal) are shared with languages of the Philippines and Sulawesi, but this just proves the conservatism of Palauan n this resect. For Chamorro, Topping ( 1 973) claimed on the basis of its verbal system, that it should e groued with Philippine languages such as Tagalog or Ilokano. His argument rests mainly on s fcus analysis of the Chamorro verbal system, which is not fully appropriate, as I will show later. Starosta and Pagotto ( 1 99 1 ) compared the Chamorro verbal system with Formosan and fcus-preserving Malaya-Polynesian languages. They note the divergent character of Chamorro, which has led them to state that Chamorro is an early offshoot from PAn. However, if Chamorro is compared with languages further south, it can e seen that this divergence results from innovations which are not eculiar to Chamorro, but are shared by most languages of Indonesia and Oceania. In this paer, I will ue evidence from verbal morphosyntax to propose a mdified subgrouping of the Malayo-Polynesian languages based on exclusively shared innovations, and establish the osition of Chamorro and Palauan within this subgrouping.2 I wll discuss the affixation of verbs in the fcus system, and in derivations from this system, and the interplay of these verbal affixes with the pronoun sets. 2 The subgrouping hyothesis proosed here is elaorated in my dissertation (in proress). The position of Chamorro and Palauan in the A ustronesian family tree 407 A note on terminology: here, antipassive and passive are defined solely as syntactic surface categories. Both are syntactically intransitive, with the agent (A) or the object (0), resectively, as subject (S), and the other participant absent, incorporated or placed in an oblique relation. This presuposes that transitive and intransitive constructions are clearly distinguishable (e.g. in person marking), which is the case for Palauan and Chamorro. 3 The basic transitive construction of both ergative and nominative languages is called 'active'. Fcus languages are treated here as neither ergative nor nominative, since it is not clear whether actor fcus (AF) or non-actor fcus (non-AF) should e the 'basic ' transitive form. 2 Reconstuction of the PMP verbal system Most Formosan and many MP languages have a verbal system usually dubed a 'Philippine-tye' focus system. In line with Blust's subgrouping, it s safe to assume that this verbal system is inherited from PAn. Ross ( 1 995) has reconstructed the focus system of PAn by concentrating on Formosan evidence, supported by additional data from MP languages. His reconstruction is summarised in Table 1 . Table 1 : PAn fcus system Past Non-past Atemoral Projective AF <umln> <um> 0 <um> -a P <I> n -u -aw LF <In> -an �n -I -ay A reduplicated form of the non-past served as a progressive. There was also a 'stative passive' expressed by *ma-I*ka-. It probably ccurred in both non-actor fcuses forming the paradigm found in Table 2 . Table 2: PAn stative passive F' LF' Past Non-past minamina- -an mama- -an Atemporal kaka- -i Projective ? ? The PMP verbal system did not differ much from the PAn verbal system (Table 3). The main innovations are the emergence of a fourth fcus and the stem extensions *paN- and *paR-. The fourth fcus using *Si- is also found in many Formosan languages (e.g. Paiwan, Atayal, Bunun); the prefix *Si- probably already existed at the PAn level as a noun forming affix (Ross 1 99 5). The stative passive remained as in PAn. 3 With this definition, it will e seen elow that nominative languages can have an antipassive (e.g. Palauan), just as ergative languages can have a passive (e.g. Chamoro). 408 Erik Zobel Table 3: The PMP verb system AF F LF IF F' LF' IF' Past Non-past Atemporal Projective <umim minaRminaN<In> <im -an ini-li- <im <um> maRmaN-en an i- ) paRpaN)/-u (?) -i -an <um> -a maR- -a maN- -a -a/aw (?) -ay ? minamina- -an minai- ma- kaka- -i a- -an ? ? ? a- -an maI- The non-atemporal forms *maN-, *maR- and *ma- are portmanteaus of *<um> plus *paN-, *paR- and *ka- with *<um>. In the following discussion, I will call all affixes containing *<um> M -afixes (including <um> itself), as opposed to base affixes (all non-AF affixes, and AF atemporal). The stem extensions *paR- and *paN- are a characteristic of the M alayo-Polynesian languages: *paR- is also found in Formosan languages, but restricted to forming reciprcals, while *paN- with nasal substitution is an innovation particular to the M alayo-Polynesian group. Although it is difficult to establish the original function of the stem extensions, it can e roughly extraolated from the mden daughter languages that in PMP *paR- had a durative and reflexive/reciprcal function, while *paN- had a distributive function, describing an action involving plural agents or objects. Both functions are transitivity­ reducing, so it s not surprising that they are found mainly in AF (see elow on the pragmatics of AF). In many Malayo-Polynesian languages, the use of the stem extensions in non-AF has een limited to fcussing circumstantial participants, such as location (LF), reason and instrument (IF) and occasionally time and manner, whereas non-AF without stem extensions fcuses core roles, such as underger and goal. These 'circumstantial' fcus forms (Table 4) are used mainly in cleft or equational constructions, which employ only the past and non-past tenses. Table 4: Post-PMP 'circumstantial' fcus forms LF" IF" Past Non-past Atemporal Projective pinaR- -an paR- -an (paR- -i) (paR- -ay) pinaN- -an paN- -an (paN- -i) (paN- -ay) ipinaRipinaN- ipaR- (paR- -an) (?) ipaN- (paN- -an) (?) I n AF, the choice of <um>, maR- and maN- has ecome lexicalised in many M alayo­ Polynesian focus languages, although some languages still allow all three forms with one verb. The position of Chamorro and Palauan in the Austronesian family tree 409 No Malayo-Polynesian language has retained the system of Table 3 completely. In particular, the atemporal and projective non-AF foms have een conflated into a single category in all Malayo-Polynesian languages. Yet both sets have to e reconstructed 'from the top down', since they can e osited for PAn and reflexes of both are found in mden M alayo-Polynesian languages. Noun case marking in PMP was as in mden Philippine languages, with nominative (marking the 'fcussed' noun phrase), genitive (ossessive, unfcused agent) and oblique (unfocused non-agent). The corresponding pronoun ets of PMP can e reconstructed as in Table 5. Table 5: PMP pronoun sets Nominative Genitive Oblique 2SG (i-)aku (i-)kau 3G a I PL.INC (i-)kita (i-)kami (i-)kamu(yu) siDa -(ng)ku -ml-nl-u -na -(n)ta -maml-nami -muyu -niDa aken iu(n) (ia ?) aen amen imuyu(n) (iDa ?) I SG I PL.EXC 2PL 3PL I n verb-initial sentences, genitive and nominative pronouns immediately follow the verb in that order (ignoring particles that also immediately follow the verb). If the sentence egins with a negative, adverb or any other memer of a class of 'preverbs', genitive and nominative pronouns are fronted to immediately follow the preverb. In such constructions, the verb usually s in the atemoral. Examples ( 1 ) and (2) from Cebuano illustrate the fronting of pronouns. 4 (1) (2) Gi-tawg-an nako siya. Pf-call-F I SG.GEN 'I called him . ' 3SG.NOM Wa nako NEG I G.GEN siya tawg-i. 3SG.NOM call- ATEMP.F 'I didn't call him.' It can e assumed that the pragmatics of fcus choice in PMP functioned as in mdern Philippine languages. Fcus selection is triggered by syntactic or pragmatic criteria. Syntactic criteria involve cases where the verb is nominalised, as in relative, existential and cleft clauses and in most WH-questions. Here, the verb must take the fcus corresponding to the function of the highlighted P. If the verb is not nominalised, fcus s triggered by pragmatic criteria: roughly seaking, if the object P is definite/referential and totally affected by the action, 4 Abbreviations used are: AT active participle, ' actor fcus, I antipassive, APPL applicative, RT article, ASP asect marker, AMP atemoral, CONJ conjunction, N genitive, GER gerundive, H UM human, IF instrument fcus, IR intransitive, lRR irealis, F lcative fcus, C lcative, G negative, OM nominative, OBJ object, OBL oblique, PAS passive, PL plural, RS ersonal article, OSS ossessive, PT past, L realis, ED reduplication, S resultative, J singular, SUB subjunctive, UF underger fcus. 410 Erik Zobel this P triggers the corresponding non-AF as in example (3) from Tagalog (if there is more than one non-agent core P, case hierarchy determines focus selection). eo anaphora for fcussed non-agents is very common in most focus languages and was certainly a feature of PMP. (3) D-in-alaw ko siya. F.PSf-invite I SG.GEN 3SG.NOM 'I visited him.' (non-AF, definite object) If the object P is indefinite, or definite but partially affected, AF is elected,s as in the Tagalog sentence (4). (4) D-um-alaw ako ng mga kaibigan . F.pSf-invite I SG.NOM GEN PL 'I visited some friends.' (AF, indefinite object) friends The syntactic trigger always overrules the pragmatic trigger, as shown in example (5), from Tagalog: on pragmatic grounds, non-AF would e selected, but since the agent is highlighted in a construction that requires nominalisation of the verb, AF is chosen. Sino ang d-um-alaw (5) a anya? who NOM F.pf-invite OBL 3SG.OBL 'Who visited him?' (AF, definite object, but agent is questioned) 3 Grammatical sketch Of Chamorro This short sketch is largely based on the descriptions by Topping ( 1 973) and Cooreman ( 1 9 8 7). Additional material is from Costenoble ( 1 940). In Topping ( 1 973), Chamorro is descried as having a fcus system of the Philippine tye. I will show elow that the notion of fcus (in the sense employed here) s not really applicable to Chamorro. Following Cooreman ( 1 987), the Chamoro verbal system is etter descried as a split-ergative system. The ergativity-split is conditioned by mod: Chamorro distinguishes realis and irrealis mood: in real is there is ergative pronoun marking, while in irrealis there is nominative marking. 3.1 Chamorro pronoun sets and verbal morphosyntax . 1. 1 Pronoun sets There are four pronoun sets in Chamorro (Table 6) with the following functions: • Set A has two slightly different subsets deending on the md of the sentence: • • • 5 the agentive Set A 1 marking A in rea lis mod; the nominative Set A2 marking S and A in irrealis mod; Set B (absolutive) pronouns mark S in realis mod and 0 in both mods; Note that definiteness of the agent is not decisive for the selection of AF. The position of Chamorro and Palauan in the Austronesian family tree • the possessive set functions as possessor and A in certain nominalisations; • free pronouns occur elsewhere. 41 1 Table 6: Chamorro pronoun sets IG 2G 3G I PL.IN I PL.X 2PL 3PL Set A2:IRR nominative Possessive Free hu (hai) hu I I -hu/ku -mu ha ta n u6 (u) ta (ai) in n n [map uh//[uma} guahu hago guiya hita hami hamyo siha Set A I : agentiv� et B absolutive yo ' hao gue ' hit ham hamyo siha -a -ta -mami -miyu -niha The ets A 1 and A2 only differ in the third erson singular and plural, and by the use of optional irrealis markers in some forms of Set A2. Both subsets of Set A are derived from the PMP genitive set (which is also directly continued in the ossessive set), while Set B and the free set reflect the PMP nominative set. The fact that the intransitive subject can e expressed by two pronoun ets makes it ossible to distinguish clearly etween syntactically transitive and intransitive constructions. 3. 1.2 Intransitive verbs Intransitive verbs can e divided into three clases deending on whether they take <um>, ma(N)- or ) as singular realis affixes. Plural pronouns with singular verb forms have dual meaning. Intransitive affixes are given in Table 7. Table 7: Intransitive verbal afixes in Chamorro Realis singular Irrealis singular ) fa(N)­ ) <um> ma(N)­ ) Realis plural maN­ manma(N)­ maN- Irrealis plural faN­ fanma(N)­ faN- The following pair illustrates the use of et A and B pronouns in irrealis (6) and realis (7) mood: (6) Irrealis: Para a aoo. IRR I pL.INC.A go 'We will go.' 6 7 Actually, u is an irrealis marker that is obligatory in the third erson and optional in the first erson plural inclusive; bai is the irrealis marker for the first erson (exclusive). ma- is historically - and probably also synchronically - identical to the passive prefix mao. 412 Erik Zobel (7) Realis: H<umanao hit. REAL.SG-go l PL. J NC B 'We went. ' . Note that the irealis forms that take Set A pronouns do not use M-affixes. . 1.3 Transitive verbs In the basic active construction, a transitive verb base is preceded by a et A pronoun marking the agent, while the object s either a pronoun of et B, as in (8) and (9), or a definite noun, as in ( 1 0). As illustrated in examples (8) and (9), active transitive forms do not change with mod, except for the slight difference in pronoun marking of the agent with ets A l (realis) and A2 (irrealis). (8) Hu-li'e' gue '. I SG.A-see 3SG.B 'I saw him.' (9) Para ai-hu-li 'e' gue'. IRR IRR- l SG.A-see 3SG.B 'I will se m.' ( 1 0) Hu-li'e' lepblo. l SG.A-see ART book 'I saw the ook . ' If the agent has to e extracted, as in cleft-, WH-clauses, relative clauses and equi-NP deletion, the agentive et A pronoun is replaced by the infix <um2>.8 The object s still represented by a Set B pronoun, as in ( 1 1 ). (1 1 ) Hayi l<umii'e' gue? who Af-ee 3SG.B 'Who saw him?' Topping has descried <um2> as an actor fcus affix. However, the latter sentence shows that the concept of focus - at least in the Philipine-tye sense - is inapplicable to Chamorro, snce transitive verbs with <um/ are Janus-faced in an dd way: to the left, the agent is highlighted, which would require AF in Philippine-tye fcus languages; to the right, the object is marked by an absolutive et B pronoun, which corresonds to non-AF in fcus languages. Avoiding the term fcus, 1 will call forms with <um2> 'active participles', following Esser ( 1 927) in his description of Mori.9 For intransitive verbs the participle is identical to the realis form of the verb. 8 9 This infix <un2, is homophonous with the intransitive infix <un , but not identical, since the latter only occurs with singular subjects. while the former is indifferent to numer. The term participle is employed. since <un2, replaces Set A person markers. which in the literature are often refered to as conjugation markers. he position of Chamorro and Palauan in the A ustronesian family tree 413 The affix maN-lfaN- is employed to form an antipassive from transitive verbs,1O i.e. a form that is syntactically intransitive with the underlying agent as subject. The antipassive s mainly used if the object is indefinite; the latter can be left unexpressed or expressed by a n unmarked noun, a s in ( 1 2) and ( 1 3). Example ( 1 2) i s i n realis mod, and the agent/subject s represented by a et B pronoun, while in ( 1 3), it is represented by a pronoun of et A preceding the irrealis variant of the antipassive prefix. The intransitive nature of the antipassive can be seen in the use of either et A or et B pronouns for the agent/subject, deending on mood, in the same manner as in examples (6) and (7) aove. ( 1 2) yo ' Man-(t)aitai lepblo. ANfI.REAL-read I SG.B book 'I read a book. ' ( 1 3) Para lepblo. bai-hulan-(t)aitai IRR IRR- I SG.A-ANTI .IRR-read book 'I will read a book. ' With certain transitive verbs, the antipassive can also ccur with definite objects, e.g. i f the object is partially affected. The object then has an oblique or lcative case marker, as in example ( 1 4). The active counterpart of ( 1 4) is sentence ( 1 5). ( 1 4) Mam-(p)atek hao gi ga 'lagu. ANTI-kick 2SG.B LOC dog 'You kicked at the dog. ' ( 1 5) Un-patek i ga 'lagu. 2SG.A-kick ART dog 'You kicked the dog. ' There are two passive affixes, <im and ma-, which are used in both realis and irrealis. Verbs with <im and ma- are syntactically intransitive, with the patient as subject. The intransitive nature of the passive is apparent in example ( 1 6), where the u nderlying patient s marked by a et A pronoun, snce it s the subject of an irrealis sentence. The agent, if present, is normally marked as oblique cae, as in ( 1 7), although in cleft and similar constructions the agent can e marked as a possessor, as illustrated in ( 1 8). ( 1 6) Ti un-h<in>engge. NEG 2SG.A-PASS-elieve 'You won't e elieved. ' (= 'He won't elieve you.') i Maria s Pedro. PASS-ee PERS Maria OBL.PERS Pedro 'Maria was sen by Pedro.' ( 1 7) L<imi'e' ( 1 8) Hafa l<in>i'e'-na 10 i Maria ? What PASS-se-3G.OS PERS Maria 'What did Maria see?' The infix (umz> instead of maN- occurs with at least two verbs, namely gimen ('drink') and the suppletive chocho ('eat' - the corresonding active base is kanno'). 414 ( 1 9) Erik Zobel Ma-li 'e ' i palao 'an. PASS-ee ART woman 'The woman was seen.' The na-passive is used when the agent is unidentified, as in example ( 1 9), or third person plural. Otherwise, the choice etween active and in-passive s deendent on rather complex discourse factors (Coreman 1 987). The transitive affixes are summarised in Table 8 . Table 8 : Transitive verbal affixation i n Chamorro Irrealis Realis Active Antipassive (SG) in-pasive ma-passive J J maN<!> ma- JaN« im) (ma-) Participle <un2> The applicative affix -i (variant form: -iyi) has several functions, one of its main functions eing the derivation of transitive stems from intransitive verbs and also sometimes from nouns. If suffixed to transitive verbs, it has enefactive function: the beneficiary then ecomes the direct object, while the original direct object is put into oblique case. The suffix -i s not a fcus affix since it can combine with all the above mentioned transitive affixes, giving maN- -i, <um2> -i, dn> -i, and ma- -i. Below I give examples for verbs suffixed with -i: hanagu-i 'go to' as-i 'pay' tugi '-i 'write tolfor' « hanao 'go') « apas 'wage') « luge ' 'write something'). Examples (20) and (2 1 ) illustrate the use of -i in a basic active clause and with the active participle affix <um2>, resectively. (20) (2 1 ) Hu-tugi '-i i Pedro i katta. I G.A-write-APPL PERS Pedro OBL letter '1 wrote the letter to Pedro.' Hu-konJotme k<um2>uentus-i ma 'gas. l SGA-agree Ar-talk-APPL ART oss '1 agree to talk to the oss.' Transitive verbs can take the suffix -(y)on to form a stative verb 'capable of being x-ed'. (Occasionally, intransitive verbs can also take this suffix to express 'capable of x'.) Some examples include: aan-on honggi-yon Jalagu-yon 'nice to look at' « atan 'look at') 'credible' « honggi 'elieve') 'capable of running' « nalagu 'run'). The circumfix JaN- -(y)an forms lcation nouns from verbs, for example: Jaiochoyan 'eating place' « chocho 'eat'). The position of Chamorro and Palauan in the A ustronesian family tree 415 3.2 Historical derivation of the Chamorro verbal system Table 9 gives an overview of how the PMP fcus affixes of Table 3 are reflected in Chamorro: Table 9 Chamorro PMP AF: Past Non-past Atemporal Non-AF: Past Non-past Atemporal *minaN- (no reflex) *maN*<um> > maN<um2> antipassive, realis mod active participle *paN- > faN- antipassive, irrealis mod *<im *<im -an > !> > (<in> -l) passive passive > > -(y)on faN- -(y)an (derives stative verbs) (derives nouns) ? > ) *-I > -i active, realis/irrealis mod active, realis/irrealis mod *-en *paN- -an . > The following developments from PMP to Chamorro should e noted: • • • • • • PMP non-AF has ecome the Chamorro active, while AF develoed into an antipassive, with the notable exception of PMP *<um>, which has tuned into the active participle, combining features of AF and non-AF. In the antipassive, the past/non-past distinction s lost, the non-past ecoming the general realis form; the atemporal has ecome irrealis mod. The non-AF atemporal has ecome the general tene form of the active, while its past form has ecome a passive; the non-AF non-past is retained residually in lexical derivations. Further, there are two innovations that are not seen in Table 9: The PMP LF atemporal suffix *-i has ecome the Chamorro applicative suffix -i which can combine with affixes that are derived from the PMP AF, maN-, <um2>, ee example (2 1 ). In the PMP atemoral, a pronominal agent with AF verbs is of the nominative et, while in non-AF it is of the genitive set, and oth are fronted to preverbal osition if the clause is oened by a negator (or any other preverbal modifier that requires the atemporal form of the verb). This is exemplified in entences (22) in AF and (2 3 ) in non-AF from Waray-Waray (Central Philippines), which has retained the patten assumed here for PMP. (22) Waray pa ako kaon. yet I G.NOM eat EG 'I haven't eaten yet.' Erik Zobel 416 kaun-a. Waray ko NEG I SG.GEN eat-ATEMP.F 'I haven't eaten [it].' (23) In contrast, sentences (24) and (25) show that in Chamorro, both in antipassive « PMP AF) and in active voice « PMP non-AF), the irrealis is preceded by et A pronouns, which are derived from the PMP genitive set. Note that the verb forms which combine with he preposed pronouns employ base afIXes (including 0). (24) Bai-hu-fa[nJ-taitai. IRR- I SG.A-ANTI.IRR-read 'I will read (something).' (25) i Bai-hu-taitai lepblo. IRR- I SG.A-read ART book 'I will read the book. ' Functionally, selection of active/antipassive in Chamorro corresponds to selection of nOJ1AF vs. AF based on pragmatic criteria in Philippine-tye focus languages. Significantly, the syntactic criteria for fcus selection in 'fcus' languages are not relevant for the selection of active/antipassive in Chamorro. This is apparent in the use of the active participle with <um]>. This is used in situations where Philippine-tye languages have syntactically conditioned AF, as exemplified in the Tagalog sentence given above in example (5). 4 Gramatical sketch of Palauan The following sketch mainly draws from two sources: Josephs ( 1 975), which contains a host of sample entences, although the analysis of data is inadequate at times; and Lemarechal ( 1 99 1 ), who has reinterpreted a gd deal of the former's analysis in a much clearer way (see the appraisal by Josephs 1 994). Additional information is taken from Patzold ( 1 968). In order to identify PMP morphemes that have been retained in Palauan, note the following sound changes: • unstressed vowels ecome J (e in Palauan orthography) or 0; • pre-stress *pa- > 0-, *pina- > ul(e)-; • Loss of *R in clusters, as in *maR- > me-, *paR- > 0- ; • *n > I, as in *<im > « i)I>. • The infx <m> « *<um» is often realised as a back semivowel or as backing of the stem vowel. 4.1 Palauan pronoun sets and verbal morphosynx Palauan has a nominative pronominal agreement system. There are five ets of pronouns or pronominal affixes, which are given in Table 1 0. The position of Chamorro and Palauan in the A ustronesian family tree 417 Table 10: Palauan pronoun sets IG 2G 3G I PL.IN I PL.X 2L 3PL.UM Free NOM I NOM II OBI OS ngak kau ngii kid kam kemiu tir ak ke ng kede aki kom te kom- -ak -au -ii -id -am -emlU -terir -k -m -I -d -(m)am -(m)iu -nr 1- dkimom1- 3PL.NON-UM ) The free, NOM I and object s are clearly derived from the PMP nominative et, while the NOM II set is related to Chamorro et A, and together with the ossessive et is derived from the PMP genitive set. The choice etween the two nominative pronoun ets depends largely on syntactic criteria: if the verb is clause-initial, the frst et is used. An exception to this are imerative sentences, where the second set is employed. If the verb is preceded by a subject (= S, A) P (or a part of it), then there is no nominative pronoun; if it s preceded by any other constituent (object, adverbial etc.), the second et s used. The second et s also obligatory after certain conjunctions. Many verb forms altenate depending on whether they are preceded by a NOM I pronoun (or a nominative P) or by a NOM II pronoun. For convenience, I will call the first verb form indicative and the latter subjunctive. Subjunctive forms never ccur without a preceding NOM II pronoun; they are also never found in sentence initial osition, except in imerative entences. The following examples illustrate the correlation etween nominative pronoun, verb affix and word order. Example (26) is a verb initial entence with a NOM I pronoun preceding an indicative verb. Examples (27) and (28) are rearranged versions of (26): in (27), the subject precedes the verb, in which case there s no nominative pronoun, while in (28), the object precedes the verb, which therefore has to e in its subjunctive form preceded by a NOM I pronoun. (26) a hong a Droteo. Ng-meng-(ch)uiu er OBL ART book ART Oroteo 3GI-ANfI-read 'Oroteo is reading the ook .' (27) a hong. A Droteo a meng-(ch)uiu er OBL ART book ART Oroteo ART ANTI-read 'Oroteo is reading the book. ' (28) A hong I-ong-(ch)uiu er ngii a Droteo. ART book 3SGII-ANTI-read OBL 3SG.REE ART Oroteo 'As for the bok, Oroteo is reading it.' Example (29) illustrates the exceptional sentence-initial position of the subjunctive with a Nom II pronoun in an imperative sentence. The non-imerative counterpart of (29) is (30). 41 8 Erik Zobel (29) D-o-rael! I PL.INCII-ITR-go 'et's go!' (30) Kede-me-rael. I PL.INC!-INTR-go 'We go. ' All transitive verbs and many intransitive verbs have two tense forms, past and non-past, and distinct forms for indicative and subjunctive. Transitive verbs can ccur in active voice, antipassive voice and three forms of passive voice. Transitive affixes of Palauan are given in Table 1 1 . Table 1 1 : Overview of basic verbal affixes in Palauan Indicative Past Non-Past Intransitive Transitive Active Antipassive Passive Resultative Gerundive ib Subjunctive Past Non-Past ib <m> me- mil- J o- ul- <m2> meNme- <ili mileNmil- J oN(me-) « i)l2> uleN(mil-) <I>, </> -el -el, -all In the active, an object pronoun is obligatory and agrees with an overt object P, as exemplified in sentence (3 1 ). By definition, a verb in the antipassive cannot take an object suffix; if the object is a pronoun or a definite P, it takes the oblique marker er. l 1 Examples ( 3 2) and (33) are antipassive entences with an indefinite and a definite object resectively. Both active and antipassive have distinct forms for past and non-past, and indicative and subjunctive. (3 1 ) Ak-kilisii kiokl. a kios-l Ak-dl2>-kios-ii lSGI-ACf-dig-3SG.OJ ART dig-GER (=hole) 'I (completely) dug the hole.' (32) a Ak-milengiis kiokl. Ak-mileN-kios ISGI-TI.PT-dig ART hole 'I was digging holes.' II Josephs descries the difference etween what is called active and antipassive here as an aspectlJal distinction etween erfective and imperfective aspect. My interpretation follows the analysis of Lemarechal, although with a different terminology. he position of Chamorro and Palauan in the Austronesian family tree 419 kiokl. Ak-milengiis er a Ak-mileN-kios I SGI - A.PST-dig OBL ART hole 'I was digging the hole.' (33) There are three passive forms: a verbal passive (me-) with two tense forms, but not distinguishing between indicative and subjunctive (34); and the resultative (35) and gerundive (36) passives, which are etter regarded as derivations outside of the transitive voice paradigm, as they are often employed as nouns. All passives generally do not allow the explicit ccurrence of the agent. 12 (34) A blai a mil-seseb. ART house ART PASS.PT -bun 'The house was bunt.' (3 5) A ulaol a ng<latech. floor ART RES-clean 'The floor is cleaned.' ART (36) A uaol a ngetach-el. floor ART clean-GER 'The floor has to e cleaned.' ART 4.2 Historical derivation of the Palauan verbal system Table 1 2 gives an overview of how the PMP fcus affixes are reflected in Palauan: Table 12: PMP fcus affixes and their Palauan reflexes Palauan PMP AF: Past *minaN­ *<umin> > > mileN<ill antipassive, past indicative active, past indicative *maN­ *<um> > meN- > mr antipassive, non-past indicative active, non-past indicative *paN- > oN- antipassive, non-past subjunctive *<im *11> *<im -an *pinaN- (-an), *(i)pinaN- > > > « i)l2> <I> </> -el active, past subjunctive (derives resultative) (derives resultative) > uleN­ antipassive, past subjunctive? Non-past *-en, *-an > -el (derives gerundive) Atemporal ? > ) active, non-past subjunctive Non-past Atemporal Non-AF: Past *-i 12 (no reflex) Josephs gives 'awkward ' examples of the verbal passive with an agent carrying the oblique marker 'er' which he susects to e based on an English mdel. 420 Erik Zobel The following developments from PMP to Palauan should to be noted: • • • • • Palauan has nominative agreement, with nominative pronouns always prefixed to the verb. The PMP nominative pronouns in post-verbal position are only preserved with object marking function. As in Chamorro, PMP non-AF has ecome active voice in Palauan, while AF developed into an antipassive, although there is 'cross-over', i.e. some active forms have an AF origin and some antipassive forms are derived from a non-AF source. Agan, we find the unusual use of 11lz>ldlz> with a following object suffix derived from the nominative t, combining features of AF and non-AF. The PMP tense distinction is preserved in Palauan, the only exception eing the non-AF non-past, which has ecome a derivational affix. Its function has been taken over by the non-AF atemporal. Among the PMP non-actor focuses, only P is preserved in the transitive paradigm of Palauan. The past subjunctive forms of antipassive voice are probably derived from LF and IF, since its function corresponds to the function of PMP 'circumstantial' non-AF fcus forms, although there is no trace of the characteristic focus affixes. The indicative uses affxes derived from PMP M-affixes, while subjunctive forms are derived from base affixes. The Nom II pronoun et is equivalent to Set A in Chamorro, and like its Chamorro counterpart, is incompatible with M-affixes. What has een said above about the functional correspondence etween non-AF vs. AF in PMP and active vs. antipassive in Chamoro, also holds for Palauan. 4.3 The -akl sux Unlike in Chamorro, there is no productive applicative suffix. However, as noted by Patzold ( 1 968), some verbs sem to contain a fossilised affix -akl. Combined with the gerundive suffix -(e)l, this gives -ekill pointing to a synchronic deep form I-akill « *-akin). There are a few pairs of verbs where the root ccurs both with and without I-akill (I-okill ·in one instance): techolb techelbakl 'wash, baptize' toir, tir tirakl 'chase' iub, ibng ibngokl -renges beko/de/rengesakl 'dive into' 'follow, oey' 'sneak out, avoid' 'sneak out, avoid' 'hear' 'having sharp hearing' (the prefix beko-, beke- means 'god at doing something'; the additional de- cannot e explained) It is very hard to extrapolate the original function of I-akill from these few examples. There are more examples where however the semantic distance of the pair is to great to exclude mere coincidence: The position oj Chamorro and Palauan in the Austronesian Jamily tree dibechakl dibech 42 1 'cross' 'invent' Most verbs that eem to bear a suffix I-akill do not occur without it, i.e. there s no way of showing that it is not an integral part of the stem: bedechakl ngeriakl techemakl techerakl 5 'throw down, drop' 'move forward' 'stuff' 'pick up with a hook' Chamorro and Palauan innovations and their occurrence in other Malayo-Polynesian languages If the innovations descried in §3.2 and §4.2 are compared, it can e seen that Chamorro and Palauan share the following innovations if compared with PMP: 1. The pronoun set A or NOM II (derived from, but distinct from the ossessive set), which ccurs efore verbs with base affixes; 2. the syntactic and semantic differentiation of <um21<m2> and maN-lmeN-, with *<um> taking over functions assciated with non-AF. Chamorro also has the following innovations not found in Palauan: 3. the circumfixes maN- -i and <um2> -i, combining PMP AF and non-AF (atemporal LF) affixes; 4. the loss of the tene (asect) distinction involving *<in>. The infix <in> has ecome a passive marker. Apart from eing a nominative language, Palauan seems to reflect one important innovation not found in Chamorro: 5. the suffix I-akill. 5.1 The Set A pronouns This pronoun et A is found in many other M alaya-Polynesian languages: Sumatran languages (including languages of the Barrier Islands), Malay, Embaloh, Old Javanese, Sulawesi languages (excluding the fcus languages of the North), and CEMP languages. Many of these languages have defective sets, e.g. Batak, Malay, Kaili. et A is not found in the Philippines, Northen Sulawesi and Boneo (except for Malayic and Tamanic), nor in Sundanese and Balinese. The most innovative feature of Set A pronouns is that they are placed efore the verb, i.e. they are prclitics or - in most languages - prefixes with a fixed osition, unlike the nominative or genitive ets in PMP or PAn, which are enclitics that are usually subject to raising. Three tyes of languages can e distinguished according to the function of the et A pronouns: 422 Erik Zobel (1 ) Set A pronouns have strictly agentive function, i.e. they are only used as agent markers with transitive verbs, as in Batak, Malay, Lampung, Embaloh, Kaili, Saluan, Toto ', Mandar. (II ) et A pronouns occur i n a split-ergative system, i.e. generally marking the agent with transitive verbs, but also marking the subject with intransitive verbs 13 in certain constructions, as in Chamorro, Buginese, Mori, Pamona, Nias. (I II) Set A pronouns are part of a nominative agreement system, not discriminating etween transitive and intransitive verbs, as in Palauan, Muna-Buton languages, and most CEMP languages. In languages of tye I, Set A pronouns co-ccur with transitive verb forms derived from the PMP non-AF atemporal. This is also the cae in languages of tyes II and III; here, additionally, intransitive verbs (including derived forms) taking et A pronouns ccur in a form derived from the PMP AF or intransitive atemporal. Usually, et A pronouns are not compatible with M -affixes. 14 Examples of Set A pronouns in languages of type I: Karo Batak (Woollams 1 996): (37) Ku-guas takal-na. I SG-thump head-3 'I clobered him on the head.' Mandar: 1 5 (3 8) U-issam-mi. I SG-know-ASP.3 'I already know.' For languages of the tyes II and II, 1 will restrict myself to glvmg examples of their ccurrence with intransitive/AF verb forms, esecially when extended with *paR-IpaN , to illustrate that these are a continuation of the PMP atemporal. Nias, Bugis and Bungku are westen MP languages, while Kamera and Buli represent CMP and EMP, resectively. - Chamorro: (39) Para bai-hu-fa-Iagu. IRR- l SG-fa- run. 'I will run.' IRR Palauan: D-o-rael! (40 ) I pL.INcII -0- go 'Let's go!' ifa- < *paR, cf. ma-Iagu) (0- < *paR-, cf. me-rael) 13 14 Including derived intransitive fonns of transitive verbs. 15 Bugis, Mandar, Bungku and Pitu Ulunna Salu data are from my own fie1dnotes. Exceptions to this are found in languages which have completely lost the F atemporal in favour of M-fonns, even in imerative function, e.g. in Toraja or Banggai. The position of Chamorro and Palauan in the Austronesian family tree Nias (Sundenann 1 9 1 3): Mi-o-fano! (4 1 ) 2PL-o go 'Go away (PL)!' - Bugis: (42) Aja ' mu-ac-cue:! Don't 2-aC-follow! 'Don't follow !' (0 - < *(p)aR-, cf. mo-fano) (a C < *(p)aR-, cf. mac-cue:) - Bungku: Nahina-po ku-pong-kaa. (43) not-yet I SG-poN-eat '1 haven't eaten yet.' poN- < *paN-, cf. mongkaa) Kamera (Klamer 1 994): u-pa-taru ? Nggiki hi (44) why CON] 2 SG-pa -watch 'Why are you watching?' pa- < *paR-) Buli (Maan 1 9 5 1 ): .. fare d-fa-pun-pun. (45 ) CON] 3PL-fa RED-hit , ...and they hit each other.' fa - < *paR-) - 423 This use of Set A pronouns with AF/intransitive verb fons is not found in languages of tye 1 . But it has to e noted that in these languages, verb fonns derived from the PMP AF or intransitive atemoral are only used as imeratives (Mandar, Totoli, Saluan), or are not reflected at all (Batak, Malay). Many languages of this tye have a defective set of preosed pronouns (Totoli, Saluan, Kaili, Embaloh, Batak). The pronoun et A could e either taken as t common innovation that ccurred in a meso-language from which all above-mentioned languages have derived, or as a n indeendent parallel innovation. The latter view is proposed by H immelmann ( 1 996) a nd Wolff ( 1 996), who regard the defective ets as incipient stages to a full paradigm. In contrast, van den Berg ( 1 996) reconstructs a full set for the parent language of at least some languages discussed here (Proto Celebic), but only for transitive verb fonns: the extended ue of set A pronouns with intransitive or AF verbs in languages of tyes II and III (both tyes being represented in his Celebic group) he regards as a later development. Here I propose that Set A is a common innovation of all the languages in which it ccurs, and that it originally was used with both non-AF and AF (and intransitive) atemporal verb fons, although the use with AF fons was more limited than with non-AF fons. Languages of tyes II and III offer ample evidence of this, esecially since they include the isolated Chamorro and Palauan languages. Only in a later, and in most cases indeendent parallel development, have languages belonging to tye I restricted the use of AF and intransitive atemoral to functions where there is usually no person marking (e.g. imperative), or have lost this atemporal completely, leading to the restriction of et A pronouns to transitive fonns derived from the PMP non-AF. Erik Zobel 424 5.2 ntipassive naN- versus active participle <un; At the end of §2 I discussed the pragmatics of AF n focus languages. Where pragmatic criteria require AF n PMP, Chamorro and Palauan have the antipassive, formed by a refle: of PMP *maN-lpaN-, while in those cases where AF is conditioned by syntactic criteria, a continuation of PMP *<um> is used (here marked <umz» . The objects of verbs carrying this <umz>, especially pronominal objects, have a case form derived from the PMP nominative. The syntactic and pragmatic differentiation of *maN- and *<umz> and the particular object marking after *<umz> are a distinctive innovation that is found in only a few areas, which however have a widely cattered distribution: the South Sulawesi, 1 6 Bungku-Tolaki and Muna-Buton groups of Sulawesi, older Toba Batak, Nias and Enggano, and probably Old Javanese and Old Balinese. I 7 In the following examples, (46) and (47) are equivalent to the Chamorro sentence ( 1 1 ), while the Nias sentences (48) and (49) parallel examples (3 1 ) and (32) from Palauan. Pitu UIunna Salu (South Sulawesi): Mea mu-ham bi-ko ? (46) mu-hit-2G who 'Who hit you?' (cf. ku-hambi-ko 'I hit you.') Bungku: (47) Iai 'umala-o ? who <um>-take-3G 'Who tok it?' (cf. ku-ala -o 'I took it.') Nias (southen dialect, Sundermann 1 9 1 3): (48) Gu-t<um>agu(-ya). I G.IRR-<um>-sew(-3G) 'I will sew it.' (definite object) (49) Gu-man-(t)agu. I G.IRR-maN-sew 'I will sew.' (indefinite object) Note that the above mentioned languages (with the exception of Old Balinese) also display innovation 1 . However, other languages of that group do not have a reflex of *<um> in the transitive paradigm, but generally use a reflex of *maN- (occasionally *maR - ). I n some cases, this can e shown to e a later development: • mdern Javanese generally has N- « *maN-) where Old Javanese used (m)aN-, (m)a­ « *maR-) and <um> (Ken 1 9 1 8-20; Zetmulder & Pedjawijatna 1 96 1 ); • in moden spoken Toba Batak, <um> is replaced by maN- (van der Tuuk 1 97 1 ); • Embaloh, shown to e closely related to Buginese, generally uses maN- (Adelaar 1 994, 1 995), with m- « *<um» only occuring efore vowels. 16 I7 In Proto South Sulawesi, *<umz' is reflected as a prefix: *um-/mu-. In some of these languages, a reflex of *maR- ccurs next to *maN- in an antipassive function, the choice of which is lexically determined. The position of Chamorro and Palauan in the A ustronesian fam ily tree 425 Based on these three examples, it is probable that the *ma[NIRJ- vs. *<um> contrast was also lost in many other languages in favour of *maN-. 5.3 The afx combinations maN- -i and (um ! -i This innovation can e descried as a symmetrisation of the fcus system. As descried aove, the PMP fcus system is asymmetrical since two or three non-actor fcuses are matched by one actor fcus. In Chamorro, the non-AF suffix -i can co-ccur with the AF affixes <un> and maN-. Thus, Chamorro -i has ecome an 'applicative' affix . I nnovation 3 is found in all languages that have the first innovation, provided they have a reflex of the affixes *naN- and *<un> (which is not the case for most CEMP languages) and the atemporal non-AF suffixes (which s not the case in Palauan). It s further found in Balinese, Madurese and Sundanese. The symmetrisation is not restricted to former LF; the AF affixes can also combine with the PMP IF atemporal *-an, giving *maN- -an and <un2> -an (e.g Totoli, South Sulawesi, Selako). Note that in all languages (except Sundanese), this innovation involves the atemoral form of PMP non-AF. 5.4 Loss of past/non-past tense distinction This innovation s found in almost all languages in which the former three innovations have taken place. An exception to this are Palauan and a few languages in a small stretch of Central Sulawesi, namely Saluan, Balantak, Kaili, and the Tomini-Tolitoli languages. These languages have retained the original PMP tense distinction. In languages that have lost the past/non-past distinction, it is the non-past form that has een lost in AF. In non-AF, some of these languages retain all tense forms, but the atemporal has ecome the general form, while past and non-past forms have een relegated to secialised meanings. This is the case for Chamorro (see §3.2), Toba Batak and Buginese. 5.5 Palauan /-aV This fossilised suffix is probably related to the widespread applicative suffix *-aken, although the vowel in the underlying final syllable presents a problem, since Palauan I! s not a regular reflex of PAn *e. However, Sirk ( 1 996) pointed out that there s much variation in reflexes of *-aken, and the Palauan form falls well within this variation. Reflexes of *-aken ccur in most languages that have at least one of the above innovations, and functionally, it has taken over the role of the atemporal IF suffix *-an. It s not found in any language that clearly did not participate in innovation 3, i.e. its intrduction must have post-dated the symmetrisation of the focus system. From its distribution, *-aken can not e reconstructed for PMP as a suffix, but it is likely that is represents the capture of an oblique-case marker *(a)ken common in the languages of the Philippines, which also survives as the Malay preosition akan. In some language groups on Sulawesi, there is evidence that this capture ccurred gradually: here, in some functions, *-aken already ccurs as an inseparable affix, while in others, *aken still etrays its originally prepositional nature (Mead 1 998). The replacement of inherited IF atemporal *-an by *(-)aken also must have ccurred gradually, with both 426 Erik Zobel morphemes occurring side by side at some stage having related but distinct functions. This s witnessed by the Malayic subgroup, where some languages have a reflex of *-an (Kendayan, Selako, some M inangkabau dialects), while others have *-aken (Malay, Serawai, Banjar). In M una, both reflexes of *-an and *-aken are found: *-an combines with second and third erson pronouns, while a continuation of *-aken is used with nouns. Functionally, the relation of the Palauan suffix /-akiV with *-aken is unclear, since there iS little agreement etween them. The major functions of *-aken are: causative (cometing with the inherited causative prefixes *pa- and *paka-), benefactive, and instrumental, none of which can e assigned to the ccurrence of /-akiV in Palauan. One has to bear in mind however that the Palauan suffix is fossilised, and a closer insection of its ccurrences might reveal a connection with the functions of *-aken listed above. At the present stage, Palauan ii at est considered an imerfect witness for the emergence of the suffix *-aken. It is even ossible that Palauan words with /-akiV reflect very early loans from an Oceanic language (cf. Proto Oceanic *-akini), such as Yaese. 6 The Nuclear Malayo-Polynesian subgroup 6.1 Reconstruction of Proto NMP The innovations discussed in §5 are not indeendent from each other, esecially the innovations proposed in §5. 1 , §5.2 and §5.3, referred to as innovations 1 , 2, 3 resectively in the following discussion: • All languages displaying innovation 2 also have innovation 1 or 3. In tun, all languages that participate in innovation 1 or 3 and which still make use of the infix <um> in the transitive paradigm, also share innovation 2 . 18 • No language that shares innovation 1 has evidence that it did not participate in innovation 3, i.e. if a language has et A pronouns, and has preserved both the AI' affixes *maN- or *<um> and the non-AF suffixes *-i or *-an, it will also make use of the innovative affix combination(s). However, a few languages that display innovation 3 do not share innovation I , such as Balinese, Sundanese and Madurese. 1 9 • I nnovations 1 and 3 are also functionally interdeendent: the use of the atemporal non·· AF suffixes in innovation 3 presupposes that in non-AF, atemporal forms are more frequent than the past and non-past forms, ecause the atemporal has taken over some of the functions of the past and non-past tenses. This is certainly the case in languages which share innovation 1 .0 This suggest that innovations 1 , 2 and 3 ccured together in a common meso-language which was a daughter language of Proto Malayo-Polynesian and from which all languages 18 19 0 An exception is Acehnese, which has prclitic pronouns that appear to e related to the Set A pronouns, but uses <eum > (from *<um» in de-transitivising function, which is quite the oposite of innovation 2. I n Sundanese and Madurese, this might e due t o the fact that pronouns have largely een relexified with nouns. This argument however does not hold for Balinese, where we still find the original P' pronouns in Old Balinese and modern Bali Aga dialects, without any evidence for Set A pronouns. This argument however cannot e applied to Balinese and Madurese. The position of Chamorro and Palauan in the Austronesian fam ily tree 427 mentioned above derived. I will call this subgroup Nuclear Malayo-Polynesian (NMP), as it contains both Malay and the Polynesian family, and the meso-language Proto Nuclear Malayo-Polynesian (PMP). (ee Srk 1 978, 1 996 for earlier attempts to use innovations 1 , 3, and 5 for reconstruction and subgrouping.) Innovation 4 (§5.4) is found in almost all NMP languages, except for those mentioned in §5.4. In one case, two closely related languages are separated by this innovation, namely Kaili (which has retained the tense distinction) and Pamona (where the distinction is lost). This shows that loss of tense is probably a drift-like phenomenon in the NMP subgroup. Innovation 5 (§5.5) also postdates PNMP, as it s only found in languages that have innovations 1 , 2 and 3, although not in all of them. As illustrated in §5.5, this innovation involved the gradual capture of the preosition *aken, eventually replacing the atemporal IF suffix *-an. We can assume that like innovation 4, this capture is the result of drift within the NMP subgroup. The reconstruction of the PNMP system for transitive verbs given in Table 1 3 accommodates the evidence given by its daughter languages. Table 13: The PNMP verb system Actor fcus Actor participle Patient fcus Gerundive Actor fcus Actor participle Patient fcus Gerundive Actor fcus Actor participle Patient fcus Gerundive Past Non-past Atempora}2 1 minaRminaN« umim <umim maRmaN<um> <um> paR­ paN0) > -en minaR- -i minaN- -i « umim -i <umim -i dm -an maR- -i maN- -i <um> -i <um> -l -i paR- -i paN- -i -i) -i n minaR- -an minaN- -an « umim -an <umim -an (i-) <im maR- -an maN- -an <um> -an <um> -an -an i- paR- -an paN- -an -an) -an This system differs from the PMP system in Table 3 in the following asects: • 21 For each patient fcus (0, -i, -an) there is a corresonding actor fcus form; this is the result of innovation 3. There was stiU a projective i n PMP, since i t is found e.g. i n Old Javanese; i t is however not relevant for the discussion here. 428 • • • Erik Zobel The patient focus non-past is formally identical to the atemporal; the PMP non-past has acquired a gerundive meaning. The active participle elongs to patient fcus, since it is followed by a patient 10 nominative case; to the left however, it highlights the agent (innovation 2). Pronominal case marking has been reshaped in the atemporal: in AF, a fronted pronoun in pivot function in PNMP is of the genitive et, not of the nominative set as in PMP, while in non-AF, only the genitive pronoun (agent) s fronted, while the nominative pronoun (object) is not fronted, schematically: PNMP: PMP: PRY aku V PRY ku V AF: PRY ku sia V P: PRY ku V sia PRY ku sia V-i PRY ku V-i sia LF: PRY ku sia V-an PRY ku V-an sia IF: Thus, in clauses with a verb n atemporal aspect, a pronominal agent is always of the genitive et, whether in AF or non-AF. Before the breakup of PNMP, these pronouns must have shifted their position from enclitic on the preverb to prclitic on the verb; ecause of their new position, they evolved into a et of their own, distinct from the genitive et, in all P daughter languages (innovation I ). The latter innovation was probably the starting oint for the development of the remaining innovations. They probably tok place in the following order (see Wolff 1 996 and Sirk 1 996 explaining the emergence of innovations 1 and 3 in a similar way): (a) (b) (c) 2 With the development of pronoun et A, AF and P atemporal ecome formally identical (although still differing with regard to the case marking of accompanying NPs). This leads to further symmetrisations of the fcus system: In AF, *<um> (or M-affixes) and et A pronouns are in complementary distribution: since *<um> ccurs if the agent is in fronted position, this use is extended to non-AF, probably first in P (Q refers to the object P in nominative case): Atemoral Agent fronted A <um>-V AF: ku-V P: ku-V Q A <um>-V o. Thus, in F the patient is in nominative case even if A is in preverbal fcus position and the verb takes *<um>, which is a major departure from the original PMP system where patients (esecially pronominal patients) of verbs with *<um> are in oblique case (innovation 2). Later this use is extended to the other non-AFs: Agent fronted Atemoral A <un>-V -i 0. ku-V -i Q LF: IF: ku-V-an 0 A <un>-V -an 0. Eventually, *-i and *-an also combine with *maN- and *maR-.2 Thus *-i and *-an ecome applicative suffixes, indeendent of focus. This co-ccurrence of AF and This is a ost-PNMP development. as in some NMP languages there is a constraint on the use of *-i/­ an with *ma[NIRJ-. e.g. in Buginese and Chamorro. he position of Chamorro and Palauan in the Austronesian fam ily tree 429 non-AF affixes leads to symmetrisation of the PMP fcus system, where originally one AF contrasts with three non-AFs (innovation 3). Two widespread phenomena are post-PNMP drifts: (d) The loss of the PMP tense distinction: former AF past forms are lost completely, while the former non-AF past tense forms *<in> and *<in> -an acquire passive function without temporal connotation (innovation 4). (e) The emergence of the applicative suffix *-aken (innovation 5). 6.2 From PNMP to Chamorro Chamorro has retained the PNMP system quite faithfully. However, some of the development descried in §3.2 are post-PNMP innovations: (i) PMPIPNMP Instrument Fcus is lost completely; (ii) *<in> -an has een replaced by <in> -i; (iii) the tense distinction etween past and non-past is lost: with intransitive verbs and in AF, only the non-past form survived as the general form in realis mood, while in non-AF, the non-past form (= the PMP atemporal) ecomes the general form for both mds, and the past form ecomes a passive; (iv) the PMP stative passive *ma-Ika- is generalised as ma-. (v) the PNMP atemporal is used for irrealis md. These developments, esecially (ii) and (iii), have also ccurred in many other MP languages, but most probably as a result of drift (loss of tense, ee §6. 1 ) or paradigmatic leveling (emergence of <in> -/.3 Incidentally, the morphosyntax of Chamorro as descried in §3 is almost identical to that found in Mori (Esser 1 927; Barsel 1 994).4 6.3 From PNMP to Palauan The case of Palauan is a little more complicated. The most significant innovation is the development of nominative agreement. Although this probably hapened under the influence of Yaese and westen Trukic isolects, with which Palauan also shares other areal features, the constructions that are involved already existed in PNMP. Indicative clauses with Nom I pronouns are derived from PMP clauses with preposed subject. For that reason, the indicative active contains the PNMP active participle infix *<um2>. Intransitive constructions with a subject pronoun following the verb fell into disuse. o. the following examples with the etyma *maR-zalan 'to walk' and *tanem 'to plant': 3 4 In eastern Central Sulawesi, Saluan has retained ,in> -an, while the related Balantak has innovative ni- -i. The similarity is a formal one: these two languages differ quite strongly in the pragmatics of the use of the passive and of the atemoral forms of intransitive verbs. 430 Erik Zobel Palauan Pre-Palauan, with fronting of subject/agent PNMP *MaR-zalan akuf. (INTR) *Aku maR-zalan. Ak-me-rael. (INTR) *MaN-(t)anem akuf. (AF) *Aku maN-(t)anem. Ak-mel-(d)alem. (ANTI) *Ku-tanem iaf. (non-AF) *Aku t<umanem ia. Ak-d<o>lem-ii. (AT) t aku and ia represent PMP nominative pronouns. More complex is the orign of the subjunctive in Palauan. As has een illustrated in Table 1 2, it is partially derived from the original atemporal, partially from non-AF forms. he original atemporal is retained in imerative and negative clauses, and after certain conjunctions; all these only use non-past subjunctive forms, as illustrated in (50) and (5 1 ). (50) D-o-rael! 1 pL.INCH -SB-go 'Let's go!' « *ta-paR-Zalan!) (5 1 ) A le-me a chull ... if 3SGH-come ART rain 'If it rains .. .' (lit. 'If the rain comes .. . ) ' In the case of fronting of non-subject constituents, the subjunctive is derived from non-AF forms of PMP, which can e seen from the fact that there is a past subjunctive form. Unlike in other NMP languages, all genitive pronouns ccurring with non-AF verbs have been fronted, not only with atemporal forms, but also with past forms, as in (52). (52) A ngikel a le-hila a bilis. fish ART 3GH-AT.PT-eat ART dog 'The fish were eaten up by the dog' ART Here, le-hil>a is derived from PMP *kdn>an-nia, with raising of the genitive pronoun. This s not found in other NMP languages, where the genitive pronoun has ecome fixed in the positionfoliowing verbs in non-AF past tense (or in the passive that has developed from it).25 As in Chamorro, the PMP stative passive *ma-/ka- is preserved in Palauan as the verbal passive prefix me-, without, however, retaining the atemoral form *ka-. 6.4 Scope and position of Nuclear Malayo-Polynesian within the Malayo­ Polynesian family If one takes the innovations discussed in 6.1 as diagnostic evidence, the NMP subgroup includes the languages of the CEMP group, Chamorro and Palauan, and most WMP languages of Malaysia and Indonesia. Not included in the NMP group are the following WMP languages: the languages of the Philippines, the three Northen Sulawesi groups 5 Compare the following examples from NMP languages: Kaili ni-kande-ku, Saluan k,imaan-ku 'J ate it', Nias ni-rongo-mi 'what you (PL) heard', as oposed to ku-kande, ku-kaan 'I will eat it' and mi-rongo 'you hear'. he position of Chamorro and Palauan in the A ustronesian family tree 431 (Gorontalo-Mongondic, Minahasan, Sangiric), the Sama-Bajau languages, Malagasy, and all languages of Boneo with the exception of the Malayic and Tamanic groups. Blust (1 999) has pointed out that his Westen group of Malayo-Polynesian (WMP) s not to e understod as a subgroup tied together by exclusively shared innovations, but just as an umbrella tem for all MP languages not included in the Central-Easten subgroup. Here, I have shown that indeed this WMP group has to e broken up by adding a node to his tre, as shown in Figure 2. Austronesian � various Formosan subgroups Mayalo-Polynesian � Philippines, north Sulawesi, north-east and nterior Boneo Nuclear Mayalo-Polynesian Westen Indonesian incl. Chamorro and Palauan Central-Easten Malayo-Polynesian Figure 2: Mdified Austronesian family tree But in tum, it has to e emphasied that the uper branches leaving the MP and P nodes are not to e taken as well-defined subgroups. There may e more than one subgroup of MP coordinate to NMP, just as 'Westen Indonesian' might contain several subgroups coordinate to CEMP. In the intrduction I mentioned that Chamorro and Palauan have no (non-trivial) common phonological innovations. Also the rammatical asects discussed here do not point to a closer relation of these two languages to each other, compared with the remaining NMP languages. What they share is a certain morphological conservatism (e.g. the retention of the maN-/<um> distinction), and the retention of the passive prefix *ma-, with loss of the atemporal form *ka-. Only a few 'Westen Indonesian' languages have preserved *ma-/ka­ as a prductive morpheme. In most languages it has een replaced by a generalised *ka­ (Javanese, Toba Batak), or the widespread innovation *taR-, which is also found in CEMP languages. Another feature that is widespread among other NMP languages but not found in Chamorro and Palauan, is the capture of the preposition *(a)ken as a enefactive and instrumental applicative suffix. Taken together, these oints indicate that Chamorro and Palauan are early offshots from PNMP. 432 7 Erik Zobel The dispersal of MP and NMP: migrations a scenario for early The family tree proposed by Blust allows some tentative conclusions about the homeland of Austronesian seakers and the way they expanded into the archipelago and Oceania. Most certainly, Taiwan must have the longest history of Austronesian ettlement; from there kers of PMP moved south to the Philippines and further to Sulawesi, Boneo, and the Sunda islands. Seakers of Proto CEMP then broke away to migrate to easten Indonesia and on to Oceania. These movements can e partly correlated with archeological findings (Bellwod 1 98 5). The NMP hypothesis allows us to refine this scenario in the following way: From the Philippines, there were movements to Borneo and Sulawesi. The latter island s a good candidate for the center of NMP disersal, since there we find a large numer of morphologically conservative languages with diverse grammatical systems. From Sulawesi, NMP seakers expanded to the Sunda islands, to parts of Boneo (as Malayic and Tamanic seakers), and to the east and south-east. It was probably also from Sulawesi that the seakers of Chamorro and Palauan (or etter: Pre-Chamorro and Pre-Palauan) sailed to the northeast to the distant islands of Palau and the Marianas. The early dates of the first settlement of these islands are in accord with the assumption that Chamorro and Palauan an! early breakaways from PNMP. Of course, these seculations do not take into account the possibility that NMP languages were also spoken in other areas, e.g. in the Philippines, and have later een replaced by fcus-retaining languages. Thus, Palauan and Chamorro seakers may have departed from an area outside of the present-day NMP area. Chamorro must have remained in relative isolation for a long time, although a few loanwords from Westen MP languages and Oceanic can e detected (Costenoble 1 940). At one oint during colonial times, the Chamorro opulation exerienced a catastrophic reduction. This and other factors resulted in massive borrowing from Spanish and, to a leser degree, from Philippine languages. Yet, its basic verbal morphosyntax sems to have remained quite unchanged. In the case of Palauan, seakers of it had contacts already in pre-colonial times with seakers of Oceanic languages, first with Yaese, then with Trukic. Ross ( 1 996) has shown that Yaese is an early offshoot of Oceanic that has borrowed heavily from Palauan and Nuclear Micronesian languages. In tum, some of the asects where Palauan deviates from PNMP might e explained by convergence with Yaese. References Adelaar, K. Alexander, 1 994, The classification of the Tamanic languages. In Tom Dutton and Darrell Tryon, eds anguage contact and chlnge in the Austronesian world, 1 -42. erlin: Mouton de Gruyter. 1 995, Problems of definiteness and ergativity in Embaloh. Oceanic Linguistics 34:375-409. Barsel, Linda A., 1 994, The verb morphology of Mori, Sulawesi. Canerra: Pacific Linguistics. ellwod, Peter, 1 985, Prehistoy of the Indo-Malaysian archipelago. Sydney: Academic rss. The position of Chamorro and Palauan in the Austronesian family tree 433 Berg, Rene van den, 1 996, The demise o f focus and the spread of conjugated verbs in Sulawesi. In Steinhauer, ed. 1 996: 89- 1 1 4. Blust, Robert A., 1 977, The Proto-Austronesian pronouns and Austronesian subgrouping: a preliminary report. University of Hawaii Working Papers in Linguistics, 9/2: 1 - 1 5. 1 993, Central and Central-Easten Malayo-Polynesian. Oceanic Linguistics 32:24 1 -293. 1 995, The osition of the Formosan languages: methd and theory in Austronesian comparative lingustics. In Li et aI. eds, 1 995 :585-650. 1 999, Subgrouping, circularity and extinction: some issues in Austronesian comparative linguistics. In Elizaeth Zeitoun and Paul Jen-kuei Li, eds Selected papers from the Eight Intenational Conference on Austronesian Linguistics. Taiei: Institute of Linguistics, Academia Sinica. Coreman, Ann M., 1 987, Transitiviy and discourse continuiy in Chamorro narratives. Berlin: Mouton de Gruyter Costenoble, H., 1 940, Die Chamoro-Sprache. 's-Gravenhage: Nijhoff. Dempwolff, Otto, 1 934-38, Vergleichende autlehre des austronesischen Wortschatzes. erlin: Reimer. Dyen, Isidore, 1 965, A lexicostatistical classication of the Austronesian languages. Intenational Journal of American Linguistics Memoir 1 9. Esser, S.1., 1 927-33, Klank- en vormleer van het Morisch. Verhandelingen van het Koninklijk Bataviaasch Genotschap van Kunsten en Wetenschappen 67, 3/4. Leiden, Bandung. Hinmelmann, Nikolaus, 1 996, Person marking and rammatical relations in Sulawesi. In Steinhauer, ed. 1 996: 1 1 5- 1 36. Josephs, Lewis S., 1 975, Palauan reference grammar. Honolulu: University Press of Hawaii. 1 994, Review of Lemarchal ( 1 99 1 ). Oceanic Linguistics 33:231 -256. Ken, H., 1 9 1 8-20, Spraakkunst van het Oudjavaansch: Verspreide Geschriten VIII + IX. 's-Gravenhage: Nijhoff. Klamer, Marian, 1 994, Kambera: a language of Easten Indonesia. Den Haag: Holland Academic Graphics. Lemarechal, Alain, 1 99 1 , Problhnes de semantique et de yntaxe en Palau. Paris: Edition du Centre National de la Recherche cientifique. Li, Paul Jen-kuei, Dah-an Ho, Ying-kuei Huang and Cheng-hwa Tsang, eds, 1 995, Austronesian studies relating to Taiwan. Symosium eries of the Institute of History and Philology, Academia Sinica 4. Taiei: Institute of History and Philology, Academia Sinica. Maan, G., 1 95 1 , Proeve van een Bulische Spaakkunst. 's-Gravenhage: Martinus Nijhoff. Mead, David, 1 998, Proto-Bungku-Tolaki: reconstruction of its phonology and aspects of its morphoyntax. PhD dissetation, Rice University, Houston. Nothofer, Bend, ed., 1 996, Reconstruction. classication, description:festschrt in honor of Isidore yen. Hamburg: Aera. Patzold, K., 1 968, Die Paau-Spache und ihre Stellung zu anderen indonesischen Sprachen. erlin: Reimer. Ross, Malcolm D., 1 995, Reconstructing Proto-Austronesian verbal morphology: evidence from Taiwan. In Li et aI., eds, 1 995:727-79 1 . 1 996, Is Yaese Oceanic? In Nothofer, ed. 1 996: 1 2 1 - 1 66. 434 Erik Zobel Sirk, 010, 1 978, Problems of high-level subgrouping in Austronesian. In S.A. Wurm and Lois Carrington, ds Second Intenational Conference on Austronesian Linguistics: proceedings, 255-273. Canberra: Pacific Linguistics. 1 996, On the history of transitive verb suffixes in the languages of westen Indonesia. In Steinhauer, d . 1 996: 1 9 1 -205. Starosta, Stanley and Louise Pagotto, 1 99 1 , The grammatical genealogy of Chamorro. In Ray Harlow, d. VICAL 2: Westen Austronesian and contact languages. Papersrom the Fth Intenational Conference on Austronesian Linguistics. Auckland: Linguistic Sciety of New Zealand. Steinhauer, Hein, d., 1 996, Papers in Austronesian linguistics No. 3 . Canberra: Pacific Linguistics. Sundermann, D. Heinr., 1 9 1 3, Niassische Sprachlehre. 's-Gravenhage: Martinus Nijhoff. Topping, Donald, 1 973, Chamorro reference grammar. Honolulu: Universitiy Press of Hawaii. Tuuk, H.N. van der, 1 97 1 , A grammar of Toba Batak. The Hague: Nijhoff. Wolff, John U., 1 996, The development of the passive verb with pronominal prefix in Westen Austronesian languages. In Nothofer, ed. 1 996: 1 5-40. W oollams, Geoff, 1 996, A grammar of Karo Batak, Sumatra. Canerra: Pacific Linguistics. Zoetmulder, PJ. and I.R. Pedjawijatna, 1 96 1 , Bahasa Parwa . Jakarta: Obor. - PRT V ­ Discussion notes Zobel, E. "The position of Chamorro and Palauan in the Austronesian family tree: evidence from verb morphosyntax". In Wouk, F. and Ross, M. editors, The history and typology of western Austronesian voice systems. PL-518:405-436. Pacific Linguistics, The Australian National University, 2001. DOI:10.15144/PL-518.405 ©2001 Pacific Linguistics and/or the author(s). Online edition licensed 2015 CC BY-SA 4.0, with permission of PL. A sealang.net/CRCL initiative.