Is Puyuma a Primary Branch of Austronesian ?

Malcolm Ross's new theory of early Austronesian phylogeny is examined. The author describes evidence that *–en served to mark verbs in undergoer voice, patient subject in a language ancestral to Puyuma, as well as evidence that * marks undergoer voice, patient subject perfective in one sociolect of Nanwang Puyuma. This evidence falsifies the claim that Puyuma reflects an early Austronesian stage at which *-en and * had not yet been reinterpreted from nominalizers into voice markers. It also falsifies the phylogeny which takes that putative innovation as its central event. A hypothetical scenario is offered to account for the replacement of the *-en, *-an and *Si(or *Sa-) series of voice markers by the series now found in Puyuma independent verbs. ha l-0 07 81 04 9, v er si on 1 In a recent paper, Malcolm Ross (2009) presents a new account of early Austronesian phylogeny, based in large part on new data from Teng's useful Puyuma grammar (Teng 2008). My intention here is to point out some problems with Ross's theory. 1 In discussing it, I will for convenience follow Ross in designating the four traditional 'focus' categories as: AV (Actor Voice), UVP (Undergoer Voice, patient subject), UVL (Undergoer Voice, location subject) and UVC (Undergoer Voice, circumstance subject). Ross's starting point is that the familiar Austronesian neutral undergoer-voice markers UVP *-en, UVL *-an and UVC *Si-, as well as the perfective aspect marker *, regarded as part of PAn verbal morphology since Wolff (1973), never occur in Puyuma, Rukai and Tsou as voice or aspect markers on verbs, but only as nominalizers on deverbal nouns. From this he infers (somewhat tentatively) that the reinterpretation of verb nominalizers into voice markers had not yet taken place in PAn. This development was assumed to be the mechanism behind the formation of the voice system of Philippine and Formosan languages in the theory of Starosta, Pawley and Reid (1982; known as 'SPQR' 2 ). Instead, Ross proposes (2009:304, 306) that this reinterpretation occurred at a later stage, and that it defines one primary branch of PAn ("Nuclear An"), which includes all An languages except Puyuma, Rukai and Tsou; each of these three representing a primary An branch (FIGURE 1). In this theory, the neutral set of UV markers found in Puyuma: UVP *-aw, UVL *-ay and UVC *-anay, occurred in that function in PAn. Ross calls them 'first-generation' affixes. The set consisting of *-en, *-an, *Siand * also existed in PAn, but according to him, only as patient, location, conveyance and perfective nominalizers respectively: he claims that they acquired voicemarking functions only in proto-Nuclear Austronesian, displacing the first-generation affixes in all but their irrealis functions. He calls them 'second-generation' affixes. Ross gives two reasons for thinking that the state of affairs represented by Teng's Puyuma must be original: first, if it were not, "we would expect Puyuma to preserve some reflex of the alleged intervening PAn stage, but it doesn’t". This means that if Puyuma was descended from a language where *-en, *-an, *Siand * had served as UV markers and not just as nominalizers, Puyuma should show traces of that usage; and second, one would have to assume that "PAn (first-generation) undergoer-voice optative/hortative forms have extended 1 Special thanks go to Sander Adelaar, Thomas Pellard, Lawrence Reid, Malcolm Ross, John Wolff and Elizabeth Zeitoun for useful discussion and information; and above all to Josiane Cauquelin for sharing with me her knowledge of Puyuma and making her notes available to me. I retain full responsibility for the final product. 2 In punning reference to the Roman senate. ha l-0 07 81 04 9, v er si on 1 their function in Puyuma to include the realis, displacing the PAn second-generation forms— a step which seems quite implausible". FIGURE 1. AUSTRONESIAN PHYLOGENY IN ROSS (2009) Ross's reason for thinking Rukai and Tsou are primary branches of An is that, like Puyuma, their verbal morphology does not make use of the undergoer-voice markers *-en, *-an and *Si(or *Sa-), or of the perfective marker *. However, alternative explanations for the absence of these affixes are available. In Tsou, auxiliary verbs expressing focus and aspect have become obligatory. Virtually every sentence has them: the lexical verbs in each sentence are the auxiliaries' dependents. Their voice markers: UVP -a, UVL -i, UVC -(n)eni probably belong to early Tsou dependent, rather than independent, verb morphology. Independent verb morphology, including the neutral UV markers, has disappeared outside of the auxiliaries. Even the auxiliaries do not take overt UV markers: an auxiliary verb belongs to undergoer voice by default, and to actor voice when it carries the AV prefix m-. As to the perfective marker *, it has been made redundant by the fact that the auxiliaries have taken over the function of indicating aspect. For Rukai, Zeitoun and Teng (in press) demonstrate that passive constructions using verbs with kihave taken over the functions of the old undergoer-voice constructions: competition from, and later replacement by, ki-passives, provides sufficient explanation for the disappearance of verb forms marked by *-en, *-an and *Si(or *Sa-). The reasons for the disappearance of * as perfective marker may be independent. Mantauran Rukai marks perfective aspect by a suffix: -nga (Zeitoun 2007:157). Ross's theory does not throw any Rukai Tsou Puyuma Nuclear An all other Austronesian languages PAn ha l-0 07 81 04 9, v er si on 1 supplementary light on the history of voice marking in Tsou or Rukai: it simply assigns to PAn certain characters which have ready explanations as post-PAn innovations. The evidence from Puyuma is therefore central to Ross's theory. Before examining whether Puyuma conforms to Ross's account, a couple of preliminary observations are in order. First, Ross's phylogeny contains a single internal node which he regards as the site of a complex innovative grammatical event. Successful phylogenies combine a number of independent innovations into a single tree. Here it must be pointed out that although the defining innovation of Ross's Nuclear Austronesian branch is complex, its components —reinterpretation of cleft sentences as verb-initial; of genitive pronouns as agentive; of nominalizers as voice markers— are crucially dependent upon one another: the SPQR mechanism can only have taken effect as a single reanalysis event, which must have occurred only once, among one group of speakers. In no way can the SPQR mechanism be regarded as a collection of independent innovations. Ross points out that his phylogeny is by and large compatible with the phonological innovations in Blust (1999); yet, because Blust's phylogeny is a ten-branch star with little higher-level structure, this point remains on the whole rather trivial. Apparently missing are compatible higher-level innovations in diffusionresistant areas like morphology and the basic vocabulary. Indeed, Ross's phylogeny clashes with several independently established innovations in these two areas: it cannot accommodate the nodes which are needed to express Tsuchida's Tsouic innovations *ramuCu 'hand' and *cani 'one' (Tsuchida 1976), 3 or the innovations proposed in Sagart (2004) for the numerals *pitu 'seven', *walu 'eight' and *Siwa 'nine'; nor can it accommodate nodes for morphological innovations like the elimination of *-en in UVP perfective verb forms, or the extension to verb roots of *kiprefixation (both innovations are discussed in Sagart, 2009) or, as we shall see below, for verbs of the shape *paR-Numeral-en. Even more problematic are facts coming from Puyuma itself: I will show in section 1 that * is still alive as a perfective aspect marker on verbs in another sociolect of Nanwang Puyuma, and that certain Puyuma verbs contain fossilized traces of the UVP marker *-en. In section 2, I will outline a hypothetical scenario for the replacement of *-en, *-an and *Si(or *Sa-) as independent UV markers by a series of markers which had until then served for 3 Ross suggests that Tsouic-only forms such as these could be retentions: if so, the widely represented *lima 'hand' and *isa 'one' (reflected in Puyuma by Lima 'hand' and sa 'one') must be post-PAn innovations displacing *ramuCu and *cani. Since however Ross's tree has Puyuma branching off directly from the root node, Puyuma should reflect *ramuCu and *cani, not *lima and *isa. ha l-0 07 81 04 9, v er si on 1 future or irrealis: *-aw, *-ay and *-anay. 1. THE FATE OF THE OLD UNDERGOER VOICE MARKERS IN PUYUMA. 1.1. The survival of *. Since Wolff (1973) it has been assumed that * was a PAn perfective marker which could co-occur on verbs simultaneously with the markers of at least three of the four voice categories, as well as derive nouns out of verbs. Ross (2009) restricts PAn * to the derivation of nominals, and perhaps of verbs in Actor Voice. The basis for Ross's claim about * can be found in Teng (2008:130). She states that Puyuma words with are nominals because they "cannot take a nominative pronoun enclitic like verbs do". Thus she takes a sentence like ku=trima na tilril ("this is the book I bought") to be a cleft sentence with two noun phrases: ku=trima 'my buying' (where ku= 'my' is a genitive pronoun proclitic) and na tilril 'the book', literally "this book is my buying". Josiane Cauquelin has worked on the same Nanwang dialect of Puyuma since the late seventies. She has these observations about the division of the village into two main social networks and the sociolects corresponding to these networks: Christianisation of the Nanwang Puyuma village began soon after World War II. All ethnic Puyuma villages have a Catholic and a Protestant Church (...). What strikes one on arrival in Nanwang is the division of the village between the two churches, which has followed the original dual structure of the village (italics mine, LS). (...) It is not possible to speak of rivalry between the two parishes but rather between these and the shamans. (...) Minor differences in language usage occasionally distinguish Protestants and Catholics. At one point I questioned elderly Puyuma about the archaic, compound personal pronoun kunu‘I (do to) you,’ as in ku-nu-beray-ay ‘I give you’; the usual form in Nanwang Puyuma is ku-beray-ay kanu. Protestants and unconverted Puyuma asserted unhesitatingly that this compound pronoun does not exist in Nanwang Puyuma, though they acknowledged its occurrence in other Puyuma dialects. Catholics, on the other hand, unanimously maintained that it exists in Nanwang Puyuma. (...) (Cauquelin, to appear). Compound proclitics are certainly ancient in Puyuma, since Tsuchida (1980:200) finds them in Tamalakaw Puyuma, as Cauquelin has pointed out to me (p.c., Nov. 2009). Teng's grammar allows only noncompound proclitic pronouns. Her principal informants are Protestants —although she has been careful to collect texts from Catholics and unconverted persons as well as Protestants, and her grammar is based on generalizations extracted from those texts. Her interactions with Puyuma speakers are conducted in Chinese or through an interpreter. In contrast, Cauquelin's informants are non-Protestants and her interactions with ha l-0 07 81 04 9, v er si on 1 Puyuma speakers are conducted in Puyuma, which gives her access to a more conversational and interactional speech level than in recorded texts or stories. The difference between Teng and Cauquelin concerning compound proclitics is probably due to a combination of sociolectal factors and factors relating to speech level. The possibility for words marked with to take a nominative pronoun enclitic appears to be another difference in the materials collected by Teng and Cauquelin: Cauquelin has collected sentences where verbs with nominative pronoun enclitics are marked for perfective by , from her informants (p.c., Oct. 2009). Examples: (1) tu=pauka=ku kana ragan m-uka i balangaw 3S.GEN=send=1S.NOM DEF.GEN priest AV-go LOC Taitung "The priest sent me to Taitung" (2) tu=piwalak=ku kan nana-li 3S.GEN=have child=1S.NOM GEN.PERS mother-1S.GEN "my mother gave birth to me" (3) piwalak=ku i puyuma have child=1S.NOM LOC Puyuma "I was born in Puyuma"