A reference grammar of Puyuma, an Austronesian language of Taiwan (review)

A reference grammar of Puyuma, an Austronesian language of Taiwan (review) Hsiu-Chuan Liao Oceanic Linguistics, Volume 50, Number 2, December 2011, pp. 590-600 (Review) Published by University of Hawai'i Press DOI: https://doi.org/10.1353/ol.2011.0025 For additional information about this article https://muse.jhu.edu/article/464591 Accessed 23 Nov 2017 22:09 GMT

Stacy Fang-Ching Teng. 2008. A reference grammar of Puyuma, an Austronesian language of Taiwan. Canberra: Pacific Linguistics 595. xviii + 309 pp. ISBN 978-085-88-3587-0. $Aust. 77.00 (Australia), $Aust. 70.00 (elsewhere), paper. Puyuma is an Austronesian language spoken by the Puyuma people living in Taitung City and Peinan Township in Taitung County of southeast Taiwan (1). Traditionally, the Puyuma are said to comprise eight villages, known as pa-fan-sher ( eight aboriginal villages ) in Mandarin Chinese. They are Puyuma (Nanwang), Katipul, Rikavung (Rikabung), Tamalakaw, Kasavakan (Kasabakan), Pinaski, Alipa, and Ulivelivek (Ulibulibuk) (3). The variety of Puyuma investigated in this study is the Nanwang dialect spoken in Nanwang and the Paoshang suburbs of Taitung City in southern Taiwan (1). Although the total number of ethnic Puyuma is 12,323 as of March 2011 (http://www.apc.gov.tw/), the number of Puyuma speakers is probably less than 1,000 (3). As of today, Puyuma is still a language with relatively few descriptions. Teng s (henceforth T) work is a welcome contribution to this under-described language. T provides a relatively clear description of various aspects of Puyuma grammar. The volume doubtless provides the starting point for more comprehensive documentation of the Puyuma language in the future. The volume under review includes sixteen chapters. Chapter 1 briefly introduces the goal of this study, the general background of the Puyuma people and the language (including its geographical setting and speakers, traditional culture and social organization, dialects, language use, orthography, and the linguistic position of Puyuma within Austronesian), previous studies, and methodology. Chapter 2, Phonetics and phonology, discusses the phonemic inventory, syllable structure, phonotactics, word stress, and morphophonemics in Nanwang Puyuma. Four points concerning Nanwang Puyuma phonology deserve special attention. First, Puyuma has a unique consonant, the voiceless retroflex stop /ʈ/, which is not found in other Formosan languages (except for Tanan Rukai, which has borrowed it from Puyuma) (12). Second, Nanwang Puyuma has voiced stops. According to Li (1991:26) and Ting (1978:325 26), Nanwang Puyuma is the only dialect of Puyuma that preserves Proto- Puyuma voiced stops as voiced stops (rather than weakening them to fricatives as in all other Puyuma dialects) (5, 12). This unique phonological feature has led to a commonly held hypothesis that Nanwang Puyuma is the most conservative dialect of Puyuma (5). Third, T claims that consonant clusters are only permitted across a syllable boundary; moreover, only nonidentical consonants or heterorganic oral stops can occur as consonant clusters across a syllable boundary (21). However, in table 2.6, we find that the following medial clusters are not possible consonant clusters in Puyuma: -ttr-, -trt-, -drt-, -dtr-, -trd-, -tdr- (21). If we follow her description of Puyuma phonotactics, we would expect these medial clusters to be possible clusters, because the two adjacent consonants are neither identical nor homorganic (one is alveolar and the other is retroflex). Similarly, we would expect sequences like - k-, -k -, -g -, - g- (that is, glottal stop next to velar stops /k/ and /g/) to be possible clusters in the language, but in fact they are not. It seems that T s description is not comprehensive enough to account for all the impermissible clusters in Oceanic Linguistics, Volume 50, no. 2 (December 2011) by University of Hawai i Press. All rights reserved.

BOOK REVIEWS 591 table 2.6. Fourth, two types of assimilation are found in Puyuma: vowel harmony and rounding assimilation. Vowel harmony, which is defined as a phonological phenomenon where neighboring vowels assimilate to each other (24), is probably better defined as a phonological process in which the quality of a vowel is altered in such a way as to make it more similar to another vowel in the same phonological word, because it applies to nonadjacent vowels. Rounding assimilation is probably better termed labial assimilation, because it is a labial (not necessarily a rounded) consonant that triggers the optional change of a schwa to a rounded vowel (25). Chapter 3, Morphology, deals with morphological units and word-formation processes (especially reduplication) in Puyuma. Morphological units identified in Puyuma are affixes, roots, stems, clitics, and words. Affixes are divided into four types: prefixes (for example, mi-, pa-, ma-, and so on), suffixes (for example, -an, -aw, -ay), infixes (<em> and <in>), and circumfixes (for example, sa- -enan) (28). Clitics are of two types: proclitics and enclitics. Proclitics include only bound genitive pronouns, whereas enclitics include bound nominative pronouns, aspectual markers (=la, =driya, and =dar), and a vocative marker =a (30). Chapter 4 deals with lexical categories in Puyuma. On the basis of morphological and syntactic criteria, two open-class categories are recognized: nouns and verbs. Forms that translate as adjectives in English are considered stative verbs rather than adjectives in Puyuma, that is, there is no distinct adjective category. Ten closed-class categories are recognized: personal pronouns, temporal nouns, deictic expressions, numerals, adverbs, tags, topic markers, aspectual markers, conjunctions, and interjections. A number of points concerning the discussion of lexical categories should be made here. First, T offers an overview of basic clause structure (section 4.2) before discussing word classes. It would have been better if this subsection were moved to either chapter 8 (Transitivity) or chapter 10 (Clause types). Second, in the discussion of verbs, T provides a classification of verbs based on valency. Whether the discussion of valency should be provided in this chapter or in chapter 8 (which deals with transitivity and ergativity) is a matter that should be reconsidered. Third, T seems to use the term deictics and demonstratives interchangeably sometimes. Thus, examples (85) (87) are referred to as locational deictics, temporal deictics, and verbal deictics in the main text, but are indicated as locational demonstrative, temporal demonstrative, and verbal demonstrative, respectively, in the free translation lines of these examples (66). Fourth, in the discussion of personal pronouns, T classifies bound pronouns into two groups: (i) a set of nominative enclitics functioning as the grammatical subject; (ii) two sets of proclitics: (a) a set of genitive proclitics functioning as the nonsubject actor in a transitive clause, and (b) a set of nominative proclitics indicating the possessor of the subject (61 62). Although the two sets of proclitics are nearly identical in form (except for the first person plural exclusive), T treats them as two unrelated sets of pronouns. Such a treatment of proclitic pronouns contradicts her claims in chapter 3 (section 3.3.1) where she identifies only a set of genitive proclitics with two syntactic functions: (a) as (an obligatory agreement marker for) the actor phrase in a transitive clause, and (b) as possessor of a possessed subject NP (30). Typologically speaking, the analysis presented in section 3.3.1 is better than the analysis presented in this chapter. Genitive pronouns typically have

592 OCEANIC LINGUISTICS, VOL. 50, NO. 2 at least two syntactic functions in Philippine-type languages: (a) they function as an actor in a transitive clause; and (b) they function as a possessor in a possessive construction. If the proclitic pronouns are considered to be a set of genitive pronouns with two syntactic functions, Puyuma genitive pronouns behave like their corresponding pronominal forms in other Philippine-type languages. The fact that an entire possessive phrase can function as a subject (but not any other grammatical relation) after procliticization of a genitive pronoun to a possessed noun can easily be accounted for by assuming that the subject phrase is unmarked for case and that what have been called nominative case markers are in fact nominal specifiers that function to identify, specify, or agree with, various features of their nominal complements (Reid 2006:9 10, Liao 2009). Furthermore, if the proclitic pronouns do not function as a possessor in a possessive phrase, there is no need to call them genitive : they can simply be called ergative. The main reason that the case form of the actor phrase in a transitive clause is called genitive in Philippine-type languages is that it is identical to the form of the possessor in a possessive construction. When such homophony does not exist, the unique case form of a transitive actor should be called ergative. Chapter 5 describes noun phrase structure in Puyuma. T begins the discussion of the topic with an outline of noun phrase structure. She uses rewrite rules and tree diagrams to present the various types of Puyuma noun phrase structures and relative clauses. However, the use of tree diagrams does not really give the reader a clear picture of the structure of noun phrases in the language. Instead, it creates unnecessary confusion (especially for readers with training in the standard X-bar module), because the tree diagrams used in the grammar do not follow from the standard X-bar module of Government and Binding Theory (nor from later versions of the theory). In a standard X-bar schema, tree structures are, in general, binary. XP is used for presenting a maximal projection or a phrasal projection (and XP = NP, VP, etc. depending on the head of the phrasal projection). However, in the grammar, XP is used for content expression, the meaning of which is not clear. Moreover, there is a node np, which does not correspond to any node level in the standard X-bar schema. Despite the confusing use of tree structures, T provides a fairly comprehensive discussion of noun phrase structures, including case, definiteness, plurality, possessive constructions (inalienable possession vs. alienable possession), NPs with a numeral or a quantifier, NPs with a demonstrative, relative clauses, and coordinate NPs. Chapter 6 deals with subject choice, mood, aspect, and morphological classes of intransitive verbs. T considers voice a notion related to subject choice, and describes Puyuma as a language that makes a four-way distinction in subject choice: actor voice (AV), patient voice (PV), locative voice (LV), and conveyance voice (CV) (109). AV clauses are considered to be intransitive, whereas non-av clauses (that is, PV, LV, and CV) are transitive (109). Two points concerning subject choice and voice deserve special attention. First, although she states that Dixon s (1994, 1997) Basic Linguistic Theory (BLT) is employed in her description, her use of the term subject is different from Dixon s (1979, 1994). In a standard BLT analysis, subject is the sole argument of a canonical intransitive construction as well as the actor phrase of a canonical transitive clause (see Liao 2004 and Daguman 2004 for details). However, subject in T s analysis only refers to the nom-

BOOK REVIEWS 593 inative phrase in both intransitive and transitive clauses, and not to the actor phrase in a transitive clause. Second, unlike most western Austronesian languages, reflexes of Proto- Austronesian (PAN) or Ross s (2009) Proto-Nuclear Austronesian (PNAN) *-en, *-an, *-in-, and *Si- are claimed to not be used in verbal clauses in Puyuma; instead, they are only used in nominalizations. This unique feature of Puyuma morphology (together with supporting evidence from Tsou and Rukai) has led Ross (2009) to propose a revised version of Blust s (1999) subgrouping hypothesis of Austronesian languages and to demote the commonly reconstructed verbal affixes *-en, *-an, *-in-, and *Si- from the PAN level to a lower node (that is, to the PNAN level). In Ross s (2009) revised Austronesian subgrouping hypothesis, PAN is assumed to have undergone a primary four-way split into Puyuma, Tsou, Rukai, and Proto-Nuclear Austronesian, from which all other Austronesian languages are assumed to have descended. Three points are of particular interest in this chapter. First, T states that in addition to the basic genitive pronominal proclitics, there is one more genitive enclitic, ti=, which is used to code desiderative mood (113). The fact that the genitive pronominal enclitic ti= can be used to code desiderative mood is interesting from a comparative perspective because a prefix ti- desiderative mood is found in Timugon Murut, an Austronesian language spoken in Sabah (Prentice 1971:134 35). Whether the similarity between Puyuma ti= and Timugon Murut ti- is the result of chance or of historical factors will be an interesting issue to be explored. Second, T reports the existence of paka- abilitative (which is used only in negative constructions) in Puyuma. Although Puyuma paka- abilitative is more restricted in both function and distribution than maka- potentive in Tagalog and other Philippine languages, it leads us to wonder whether paka- and/or maka- can be reconstructed to the PAN level. Third, Zeitoun and Huang (2000) argue that the prefix ka- indicates stativity in nonfinite constructions in a number of Formosan languages. In contrast, T notes that in Puyuma it is not unusual for ka- to cooccur with verbs that are semantically dynamic (123). Chapter 7, Transcategorial operations, deals with verbalization and nominalization. Verbalization is an operation that forms a verb from a nonverbal element, whereas nominalization is an operation that forms a noun from a nonnominal element. Nominalizations are analyzed as either lexicalized nominalizations or gerundive nominalizations (129). Lexicalized nominalizations differ from gerundive nominalizations in that the former are nonproductive and are negated like a nominal construction, whereas the latter are productive and are negated like a verbal construction (129 30). T divides lexical nominalizations into six groups depending on the semantic functions of the nouns resulting from the nominalizing operations: 1 (i) action/state nouns; (ii) person-denoting nouns; (iii) patient nouns; (iv) instrumental nouns; (v) locative nouns; and (vi) temporal nouns. Chapter 8 deals with case marking, transitivity, and ergativity. Puyuma makes a threeway case distinction (nominative, genitive, and oblique) in its personal pronouns, but a two-way case distinction (nominative and oblique) in its full NPs. One thing that deserves special attention is the status of oblique NPs. In a number of published materials, an oblique NP is commonly considered to be an adjunct and not an argument. However, the 1. T uses two similar terms, lexicalized nominalizations and lexical nominalizations. It is not clear from her description whether or not these refer to the same thing.

594 OCEANIC LINGUISTICS, VOL. 50, NO. 2 author, concurring with Liao (2004), considers an oblique NP to be either an argument (for example, the indefinite patient required by a bivalent intransitive verb) or an adjunct, depending on whether it is required by the valency of a verb. If it is required by the valency of a verb, then it is an argument; if not, it is an adjunct. T recognizes three types of verb in Puyuma: ambient, intransitive, and transitive. A transitive verb is a verb that takes two core arguments, an intransitive verb is a verb that takes one core argument, and an ambient verb is a verb that does not take any argument. No ditransitive verbs are recognized in Puyuma because the maximal number of core arguments that a Puyuma verb can take is two (144). Strictly speaking, the use of the term ambient in the discussion of transitivity is inappropriate and should be replaced by atransitive (see Daguman 2004:111), because ambient is a valency-related notion (56) rather than a transitivity-related notion. In the discussion of ergativity, T claims that Puyuma displays syntactic ergativity because both S (the sole argument of an intransitive clause) and O (the undergoer of a transitive clause) are marked as nominative, while A (the actor of a transitive clause) is marked differently from both S and O (162). However, she claims that Puyuma verbal morphology is accusatively aligned because antipasssive verbs have the same marking as intransitive verbs (162). T s discussion of ergativity in Puyuma is questionable in three senses. First, ergativity is a notion that is supposed to apply to case-marking and/ or agreement (see Dixon 1979 and 1994), but not to verbal morphology. Second, in studies that distinguish morphological ergativity from syntactic ergativity, morphological ergativity refers to the S/O alignment pattern of nominal case-markers and/ or agreement forms, but syntactic ergativity refers to the S/O alignment pattern of syntactic processes (for example, relativization, quantifier floating, and so on). What is referred to as syntactic ergativity in this study would be better referred to as morphological ergativity. Third, even if antipasssive verbs have the same marking as other intransitive verbs, this would not suggest that Puyuma verbal morphology is accusatively aligned, because T has not shown that the oblique NP of such constructions carries accusative case-marking. Chapter 9, Re-encoding of arguments, deals with processes that are commonly referred to as valency-changing, including causative constructions, reciprocal constructions, reflexive constructions, anticausatives, and ki- passives. Two points deserve further discussion. First, reciprocal constructions are marked by either of the reciprocal prefixes mar(e)- (for stative verbs) or ma-ca-/ ma-cvcv- (for dynamic verbs) (172 74). This agrees with Zeitoun s (2002, 2010) reconstructions of PAN reciprocals: *mar- for stative verbs, but *maca- for dynamic verbs. However, reciprocal prefixes are used not only to express reciprocality, they can also be used in the following situations: (i) repeated occurrences of a situation; (ii) a collective event; (iii) an increasing degree (that is, more and more ~ ); (iv) a chaining situation; and (v) a distributive situation. It would be interesting to see whether any of the nonreciprocal uses of *mar- in Puyuma agree with those of reflexes of *mar- in Malayo-Polynesian languages. Second, T distinguishes the anticausative prefix mu- from the prefix mu- used in motion verbs (179, 180). She considers the former to be monomorphemic, but the latter bimorphemic, consisting of m- intransitive plus u- motion (179).

BOOK REVIEWS 595 Chapter 10, Clause types, introduces three types of clause: (i) verbal clauses, (ii) nominal clauses, and (iii) locative/existential/possessive clauses. Verbal clauses are divided into three groups: transitive clauses that take two core arguments, intransitive clauses that take one core argument, and ambient clauses that do not take any argument. Transitive clauses are either bivalent or trivalent; a trivalent clause is considered to be an extended transitive clause rather than a ditransitive construction because it takes two core arguments and one oblique argument (rather than three core arguments). Intransitive clauses are either monovalent or bivalent. Monovalent clauses can denote either stative or dynamic events. Bivalent intransitive clauses are composed of three groups: (i) extended intransitives that take an actor argument and an oblique argument, (ii) anticausatives that take an undergoer argument (and an oblique argument), and (iii) passives that take an undergoer argument (and an oblique argument) (188 90). Although both anticausatives and passives take an undergoer subject, they differ in that the undergoer/patient subject of an anticausative is nonvolitional, whereas the undergoer/patient subject of a passive exercises some degree of intention (188, 190). Two problems are observed in T s discussion of intransitive clauses. First, she considers stative and dynamic monovalent predicates as having the same argument structure (189). Such a statement is problematic in that a stative predicate takes an undergoer argument, whereas a dynamic predicate takes an actor argument. More specifically, in the verb s argument grid, one contains an undergoer whereas the other contains an actor. Thus, stative and dynamic monovalent predicates, although they have the same number of arguments, do not have the same argument structure. Second, T classifies anticausatives and passives as bivalent verbs. However, some of the examples that she provides in the grammar are clearly monovalent for instance, examples (11) and (12) on p. 190. Moreover, cross-linguistically, agentless passives are much more frequent than passives with an agent. Whether anticausatives and passives should be considered bivalent or monovalent is a matter that needs to be reconsidered. Nominal clauses are of two types: classifying clauses and identifying clauses. Although both types of clause require a nominal predicate, they differ in two respects. First, the predicate nominal that occurs in classifying clauses is indefinite, whereas its counterpart in identifying clauses is definite (191 92). Second, an optional copula verb amau or its negative counterpart ameli can occur in identifying clauses, but not in classifying clauses (191 92). Whether amau and ameli should be considered full-fledged (negative) copulas is a matter that deserves further investigation. Chapter 11, Negative constructions, recognizes three types of negative clause: (i) negative verbal clauses, (ii) negative nominal clauses, and (iii) negative locative/existential/possessive clauses. Negative verbal constructions are introduced by the negative particle adri, negative nominal clauses by the negative verb ameli, and negative locative/existential/possessive clauses by the negative verb unian. In addition to adri, ameli, and unian, there is a fourth negative element, maulrid not know, which is used as a negative counterpart of maladram know; understand. One thing to be noticed is that T analyzes ameli, unian, and maladram as verbs, but adri as a particle. She offers evidence to argue against adri as a noun, verb, or adverb (209 10) and concludes that adri forms a morpheme category of its own (210). One piece of evidence that she uses in

596 OCEANIC LINGUISTICS, VOL. 50, NO. 2 arguing against adri as a verb is that adri does not change its form for different aspects or moods (210). However, in chapter 14, we find that adri can be suffixed by the projective mood marker a, as in adri-a drua na ala (I/We/They wish) the enemies won t come. ; cf. maruwa-a=ku m-ulra esi (I wish) I can succeed. (259). If adri is not a verb, how can we explain the occurrence of the projective mood marker on it? Moreover, in chapter 14, we also find that adri can be (causativized and) nominalized: for example, tu=rengarengay-aw=ku dra pa-ka-adri-an tr<em>ekelr dra eraw He persuaded me not to drink wine. ; cf. tu=rengarengay-aw=ku dra (pa-)tra-trekelr-an dra eraw He persuaded me to drink wine. If adri is simply a particle, as T concludes, the question arises as to whether cross-linguistically, can particles be nominalized and/or causativized? Based on the facts that adri can be suffixed by the projective marker -a and be causativized (pa-) and then nominalized (-an), it would seem that adri is better analyzed as a verb (possibly a stative verb because of the existence of the prefix ka-) rather than as a particle. Chapter 12, Non-declarative clause types, deals with imperative constructions, interrogative constructions, and hortative constructions. Generally speaking, imperative constructions are used to give commands and instructions or to make requests. In Puyuma, affirmative imperative sentences are also used when the speaker is of a higher rank than the addressee (217, 219). However, when politeness is considered, two types of declarative sentence are used. instead, to make requests (219). In one type of declarative sentence, the addressee is overtly expressed by a second person pronoun; in the other, the addressee is expressed by the first person inclusive pronoun ta= or =ta (218). The first type of declarative sentence is used when the addressee is of a higher rank than the speaker, whereas the second is not restricted to such situations (219). As for the negative imperative, or prohibitive, construction, it does not observe such a politeness restriction. Interrogative constructions are of three types: yes/no questions, alternative questions, and information (or question-word) questions. Yes/no questions have two subtypes: neutral yes/no questions and biased yes/no questions. Alternative questions are usually formed by juxtaposing two alternatives and optionally connecting them with andri if not (possibly from an adri when/if not ) (223). Information questions are indicated by the use of interrogative proforms and rising-falling intonation. Puyuma interrogative proforms are of four categories: nominal, adverbial, verbal, and numeral (224). Information questions (at least nominal interrogatives) appear to exhibit a wh- in situ pattern in Puyuma (224). This appears to be quite different from Tagalog and other western Austronesian languages that require the use of a wh- pseudo-cleft pattern when a nominative phrase is questioned. Hortative constructions are usually formed by suffixing the projective suffix -a and/or by cliticizing the first person inclusive pronoun ta=/=ta (227). Chapter 13 deals with serial verb constructions (SVCs) in Puyuma. Following Crowley (2002:10), T defines SVCs as syntactic constructions involving what can be analyzed at the surface level as single clauses, but which are nevertheless expressed by means of multiple predicates (229). In Puyuma, initial verbs in an SVC can be either transitive or intransitive, but noninitial verbs in an SVC can only be intransitive (230). Irrealis mood can only be marked on the first verb (V1) but not on the second verb (V2). Similarly, in imperative con-

BOOK REVIEWS 597 structions, only V1 can appear in the imperative form. As for negation, the negator adri can never precede V2 in an SVC. The shared argument can be manifested as either a clitic pronoun or a full NP. If the shared argument is expressed as a clitic pronoun, nominative or genitive, it can only attach to the first verb (231). By contrast, if the shared argument is expressed as a full NP, the position of the shared argument differs from construction to construction. In a same-subject (Type I or Type II) SVC, it can appear either between V1 and V2 or after V2. However, in a switch-subject SVC, the shared argument (that is, the oblique-marked actor) must appear before V2 (232). As for an inclusory SVC, the shared argument can appear after V2. Chapter 14, Complement clauses, deals not only with complement clauses, as suggested by the title, but also with other complementation strategies (that is, nominalization and SVC). In Puyuma, complement clauses can only function as oblique arguments, but never as S, A, or O arguments (249), and are always introduced by dra, an indefinite oblique noun phrase marker (248-249). Morphosyntactically, verbs in complement clauses behave like verbs in independent clauses. More specifically, they can be either transitive or intransitive, depending on whether there is a definite undergoer: if there is a definite undergoer, they will be transitive; if not, they will be intransitive (249). Moreover, they are free to have their own aspect and mood. Thus, even if the matrix verb is in a realis form, the verb of an embedded complement clause can be in an irrealis form. Furthermore, negation is allowed in a complement clause (250). Chapter 15, Adverbial clauses, introduces three types of adverbial clause: (i) clauses denoting temporal relations, (ii) reason and result clauses, and (iii) conditional clauses. Clauses denoting temporal relations are of four types: (i) clauses denoting temporal posteriority ( before clauses); (ii) clauses denoting temporal anteriority ( after clauses); (iii) clauses denoting a temporal boundary ( since and until clauses); and (iv) clauses denoting temporal overlap ( when and while clauses). Reason and result clauses are marked by the topic marker i; whether an adverbial clause may convey a reason or a result depends on the context. Conditional clauses are of two types: (i) conditional clauses denoting real events, those that refer to real present, habitual/generic, or past situations (269); and (ii) conditional clauses denoting unreal events, which are indicated by an irrealis marker (usually Ca- reduplication, ka- marking, or affixation of <a> on the verb) (269). Chapter 16, Coordination, deals with three types of coordination: (i) coordination of noun phrases, (ii) coordination of clauses (including symmetrical coordination, asymmetrical coordination, and ellipsis in clausal coordination), and (iii) adversative coordination. Coordinate noun phrases are generally encoded by the same case form, although occasionally they might be encoded by different case forms (275). In the case of coordination of noun phrases, the linear order of the coordinands is irrelevant in interpreting the meaning. However, in the case of coordination of clauses, the linear order of coordinands might be crucial in interpreting the meaning. Based on the reversibility of the coordinands, we can divide coordination of clauses into two types: (i) symmetrical coordination, in which the reversal of the linear order of the coordinands has no semantic significance, and (ii) asymmetrical coordination, in which reversing their order will cause

598 OCEANIC LINGUISTICS, VOL. 50, NO. 2 a change in meaning (277). Adversative coordination is expressed by amuna but, which can only connect clauses (279). The grammar concludes with three appendices: (i) a list of 26 transcribed texts (with title, speakers, genre, and approximate length indicated) used in the grammar; (ii) a list of 45 affixes in Puyuma, their meanings, a few carefully chosen examples of their use, and a cross-reference to the section of the grammar where they are described; and (iii) two texts (one narrative and one procedural), with interlinear and free translations. A list of references completes the work. In general, there are three types of problem found in the grammar: clarity, consistency, and organization. Some examples with problems in clarity: (i) In Papulu, Taiwanese seems to be used more often than Mandarin (6). The author does not mention where Papulu is and we do not find Papulu in map 2 either. (ii) In section 2.1 (an overview of syllable structure), the author states: Any of the four vowels (11). However, those four vowels are not introduced until section 2.2.2. (iii) The main discussion of serial verb constructions (SVC) is in chapter 13, but the abbreviation SVC is used in chapter 3. The author does not include SVC in the list of abbreviations, nor does she mention what SVC stands for when it is first used on p. 33. Some examples with consistency problems: (i) Pronouns that are used in possessive constructions are sometimes referred to as genitive pronouns (30), but sometimes as nominative pronouns (61 62). (ii) The form ulaya is sometimes referred to as a copula verb (193, 195, 196), but sometimes as an existential verb (198). (iii) The form ameli is sometimes analyzed as a form of analytic negation (206), but sometimes as a form of lexical negation (214). (iv) T states that three types of question are distinguished in Puyuma: yes/no questions, alternative questions, and information or questionword questions. Their common feature is that they all have a rising-falling pattern of intonation (220). However, in the discussion of biased yes/no questions, she states that phonologically, a biased yes/no question ends with rising intonations (222). (v) The gloss to face a certain direction sometimes takes the form pia- (129, 231), but sometimes takes the form piya- (242, 266, 283). Some reorganization of the grammar would have been helpful. (i) In chapter 2, the order of section 2.2 (phonemic inventory) and section 2.1 (an overview of syllable structure) would have been better reversed. Moreover, section 2.1 could have been merged with section 2.3 (on the syllable). (ii) Chapter 10 (clause types), especially the discussion of nominal and verbal clauses, would have been better earlier in the grammar. A variety of topics related to the discussion of verbal clauses, which are distributed in various chapters throughout the book, would have been better incorporated into the chapter that deals with verbal clauses. These include such topics as subject choice, mood, and aspect (chapter 6), transitivity (chapter 8), valency (chapter 4), and the discussion of imperative constructions (section 12.2). In addition to the problems mentioned above, a number of typographical errors are found in the volume, including the use of advice for advised (69), makatelun mi-sama for makatelrun mi-sama (74), use for used (113), 13.4.2 for 8.4.1.4 (257). More serious mistakes are found in two statements concerning the general background of the Puyuma people. More specifically, T states that they dominated

BOOK REVIEWS 599 eastern Taiwan during the period when the Chin Dynasty and then Japan ruled Taiwan in the 18th and 19th centuries (2). This statement should be read as they dominated eastern Taiwan during the period when the Ching Dynasty and then Japan ruled Taiwan in the 19th and 20th centuries. Despite the critical comments in this review, this volume provides a very accessible description of Nanwang Puyuma structures, with carefully prepared summary statements and charts appearing in every chapter. Most, if not all, analyses are well described and illustrated with numerous examples, providing a rich resource for comparative studies of Formosan languages. A number of insightful observations and interesting data (for example, the discussion of the paka- abilitative construction, the nonreciprocal functions of reflexes of *mar-, various uses of ka-, and so on), which are significant for future studies of comparative Austronesian morphosyntax, are provided. The volume is definitely an important piece of work that linguists interested in Austronesian languages, language typology, and language documentation would like to keep on their bookshelf. HSIU-CHUAN LIAO National Tsing Hua University REFERENCES Blust, Robert. 1999. Subgrouping, circularity and extinction: Some issues in Austronesian comparative linguistics. In Selected papers from the Eighth International Conference on Austronesian Linguistics, ed. by Elizabeth Zeitoun and Paul Jen-kuei Li, 31 94. Symposium Series of the Institute of Linguistics (Preparatory Office), Academia Sinica, Number 1. Taipei: Academia Sinica. Crowley, Terry. 2002. Serial verbs in Oceanic: A descriptive typology. Oxford: Oxford University Press. Daguman, Josephine Sanicas. 2004. A grammar of Northern Subanen. PhD diss., La Trobe University. Dixon, R. M. W. 1979. Ergativity. Language 55(1):59 138.. 1994. Ergativity. Cambridge: Cambridge University Press. [Cambridge Studies in Linguistics No. 69.]. 1997. The rise and fall of languages. Cambridge: Cambridge University Press. Li, Paul Jen-kuei. 1991. Orthographic systems for Formosan languages. Taipei: Ministry of Education. Liao, Hsiu-chuan. 2004. Transitivity and ergativity in Formosan and Philippine languages. PhD diss., University of Hawai i at Mānoa.. 2009. On the development of the nominal phrase marking system in Ilongot. Paper read at the Conference on South East Asian Languages, Université Paris Diderot and INALCO, Paris, France, December 17 19. Prentice, D. J. 1971. The Murut languages of Sabah. Canberra: Pacific Linguistics. Reid, Lawrence A. 2006. On reconstructing the morphosyntax of Proto-Northern Luzon. Philippine Journal of Linguistics 37(1):1 64. Ross, Malcolm. 2009. Proto Austronesian verbal morphology: A reappraisal. In Austronesian historical linguistics and culture history: A Festschrift for Robert Blust, ed. by Alexander Adelaar and Andrew Pawley, 295 326. Canberra: Pacific Linguistics.

600 OCEANIC LINGUISTICS, VOL. 50, NO. 2 Ting, Pang-hsin. 1978. Reconstruction of Proto-Puyuma phonology. Bulletin of the Institute of History and Philology, Academia Sinica 49:321 92. [In Chinese.] Zeitoun, Elizabeth. 2002. Reciprocals in Formosan languages. Paper presented at the 9th International Conference on Austronesian Linguistics (9-ICAL), Australian National University, Canberra, Australia, January 8 11.. 2010. On the reconstruction of reciprocal prefixes in PAN based on Formosan data. Guest Lecture: LING581500: Comparative Austronesian morphosyntax, Graduate Institute of Linguistics, National Tsing Hua University, Hsinchu, Taiwan, November 24. Zeitoun, Elizabeth, and Lillian M. Huang. 2000. Concerning ka-, an overlooked marker of verbal derivation in Formosan languages. Oceanic Linguistics 39:391 414.