Clause structure and adjuncts in Austronesian languages (Review)

Just as teenagers divide the world into jocks, nerds, goths, preps, townies, posers, and so on, linguists tend to categorize languages in broad strokes by what is best known about them: Bantu languages are defined by applicatives and tones, Slavic languages by palatalization and aspect, Native American languages by polysynthesis and attrition, and so on. Ask a linguist who does not work on Austronesian what he or she knows about these languages and you are likely to hear about nasal substitution, infixation, verb-initial word order, articulated systems of grammatical voice, or restrictions on movement. Any book that is addressed to the general public, not just to Austronesianists, and that takes the reader beyond these staple topics fills an important gap. The book reviewed here does just that and more. If pressed to name its goals, I would identify them as twofold: to make issues that at first sight may seem parochial to Austronesianists relevant to the entire field by identifying the theoretical challenges they pose, and to move the analysis of Austronesian beyond the traditional stereotypes, such as named above, but in a way that also connects the new theoretical challenges with established ones. The interests of the three editors are nicely distributed: Gärtner is best known for his semantic work on a number of languages, Sabel is a dedicated explorer of everything Malagasy who has done important work in theoretical syntax, and Law combines interests in syntax with attention to Malagasy. Leaving phonology and morphology aside, the most publicized, although not yet fully understood, aspects of Austronesian languages include verb-initial word orders (often accompanied by the right-hand subject), articulated voice systems, and very rigid constraints on what can be extracted—usually the only constituent undergoing extraction is subject-like, or what many researchers call an external argument (“subject-only restriction”). The introductory chapter, “Clause structure and adjuncts in Austronesian languages,” by Gärtner, Law, and Sabel, does an excellent job surveying the main approaches to these three phenomena, and in so doing will be of clear use not only to specialists in Austronesian but also to generalists and syntacticians. With respect to verb-initiality, a number of researchers have argued that pathways to verb-initial order may differ across languages, and even within a language family or group. The well-recognized possibilities include (i) nonconfigurationality (see, for example, Austin 2004); (ii) base-generation, with parametric variation in the position of the specifier in functional projections either to the right or to the left (the rightward specifier of T allows one to derive VOS order in a direct way); (iii) subject lowering (as argued for Chamorro and Māori by Chung 1998); (iv) verb-raising; and (v) VP-raising. The introductory chapter does a particularly good job presenting options (iv) and (v) and discussing their pros and cons. Still, more explicit comparisons with other verb-initial languages outside the Austronesian family would have been useful—after all, these languages may present additional test cases for particular theories. This criticism notwithstanding, the background part of the introductory chapter is clearly written and can be a valuable companion in courses on verb-initiality, restrictions on

ful-after all, these languages may present additional test cases for particular theories. This criticism notwithstanding, the background part of the introductory chapter is clearly written and can be a valuable companion in courses on verb-initiality, restrictions on A-bar movement (the subject-only condition) and complex verbal morphology. The editors also show that each of the three topics has been subject to intense debate, and present varying perspectives in a careful, thoughtful manner.
The editors then build on their survey of the main approaches to verb-initiality, complex voice systems, and the subject-only restriction to introduce two main issues in the syntax of adjuncts as viewed from the vantage point of Austronesian: theories of adjunct licensing and placement, and argument vs. adjunct asymmetries in extraction.
Word order derivation is central to the topic of licensing adjuncts and their position in a clause. The introductory chapter and several of the following chapters address two competing approaches to licensing and linearization: the feature-based approach and the semantic approach. The feature-based approach, rooted in Kayne's influential idea that syntax allows only left-branching (Kayne 1994), finds its most articulated expression in Cinque's proposal that adverbs are specifiers of left-branching functional projections. As such, each adverb is associated with a particular functional head (Cinque 1999, and also Alexiadou 1997, Laenzlinger 1998. On this approach, the placement of a particular adverb in two different positions in a clause signals that it is licensed by different heads; accordingly, different positions for the same adverb are expected to yield different interpretations, as in the following: (1) a. These rooms are always intentionally closed to the public.
b. These rooms are intentionally always closed to the public.
The alternative "semantic" approach, articulated by Thomas Ernst (Ernst 2002(Ernst , 2007, and also Svenonius 2002, Shaer 2003, states that adverbs can adjoin to the right or to the left (no antisymmetry); they can adjoin to either maximal projections or to X-bar; and they are licensed semantically. The semantic licensing arises from the lexical-semantic properties of individual adverbs (that is, does the adverb scope over an event, a proposition, etc.?) together with the compositional semantics of the entire clause (that is, does the adverb fit in with the layered event structure?). On this account, multiple linearization possibilities do not necessarily yield different interpretations because it does not posit different positions in the clause spine. The following English data show that this prediction is correct: (2) a. She frequently would have been visiting Sam.
b. She would frequently have been visiting Sam. c. She would have frequently been visiting Sam. (Ernst 2007: ex. [23]) The two competing accounts make different predictions for verb-initial languages. On the syntactic, feature-based approach, the ordering of adjuncts in the preverbal and postverbal domain must be derived differently. By hypothesis, preverbal adjuncts are basegenerated, and they should therefore follow the order proposed by Cinque. This seems to be confirmed by Malagasy data, as examined by Rackowski and Travis (2000) and reevaluated in this volume by Thiersch. In Malagasy, preverbal adverbs follow the universal order (I am using English equivalents only): (3) generally > already > always The postverbal order of adverbs in Malagasy seems to be the mirror image of what is claimed by feature-based accounts. If one were to maintain the feature-based account of adverbial placement, it would necessitate a particular approach to deriving verb-initiality: the verb would have to undergo cyclic raising (intraposition), and each of the raising operations would leave a stranded adverb in its wake. This approach is possible (per Rackowski and Travis 2000), but it introduces the need for additional evidence of such verb-raising; several of the chapters in this collection confront this issue.
The other important issue addressed by the editors in the introductory chapter has to do with the asymmetry between arguments and adjuncts with respect to A-bar movement. Austronesian languages are known for the following constraint on extraction: the only DP argument that can undergo topicalization, relativization, or wh-movement (if it is available-see below) is the most prominent DP, which has been variously characterized as a topic, pivot, external argument, trigger, or subject. This restriction is widespread in Austronesian. It occurs, for instance, in Malagasy, most Philippine languages, Formosan languages, languages of Indonesia, and Polynesian languages. In some of the transparently ergative languages, such as Tongan or Samoan, it resurfaces as absolutive-only extraction. Adjuncts, however, seem to avoid this constraint: Chamorro, Malagasy, Indonesian, Sasak, and Māori appear to allow some adjuncts to extract freely without any special operations that promote them to the status of external argument. For example, in Chamorro, where argument extraction is always accompanied by special inflectional morphology (wh-agreement as described by Chung 1998), possessors of direct objects surface without any special marking on the verbs (Chung 1998:255); also, when a higher adjunct (S-adjuncts) forms an unbounded dependency, the higher verb does not show any wh-agreement (Chung 1998:363-65). In Tagalog, PPs which are not subjects/external arguments can extract across a clause without any special morphology on the predicates of the intervening clauses (Kroeger 1992:219-20). In Polynesian languages, locative, temporal and causal adjuncts often topicalize without a dedicated marker (ko, 'o) which is required under the extraction of arguments (cf. Moyse-Faurie 2004).
It would be fair to say that our ability to explain extraction in Austronesian in general and extraction of adjuncts in particular is still in the initial stages of discovery. Relating the subject-only constraint to the general issues of A-bar movement makes Austronesian languages particularly intriguing. The introductory chapter and the material in other chapters of this book advance our knowledge by pointing out that not all adjuncts are equal: it seems that the extraction of adjuncts is sensitive to the distinction between PPs and adverb-like adjuncts (the former are more likely to extract), and to the different semantic categories of adjuncts. The privileged classes include adjuncts that modify the entire event: temporal, locative, and sometimes purpose and cause expressions. Bringing this issue to the attention of scholars working on Austronesian may help identify more specific classes of constituents that undergo extraction without special marking.
The introductory chapter is followed by seven chapters which cover different Austronesian languages from Seediq, Tagalog, and Malagasy in the western branches of the family, to Niuean in the Polynesian group of Oceanic. Although the book is certainly Western Malayo-Polynesian heavy, many of the issues raised for those languages could and should also be revisited from the perspective of Oceanic.
The authors are well-known to readers interested in Austronesian, and several have also done important work in theoretical syntax; this combination of general syntactic interests and empirical familiarity with Austronesian is what makes this collection particularly noteworthy.
The chapter by Craig Thiersch, "Three systems of Remnant Movement II and extraction for specifier position," provides a welcome addition to the overview of the approaches to linearization and the subject-only restriction discussed in the introductory chapter. Thiersch starts from a Kaynian antisymmetric approach, which necessitates the conception of remnant movement as a means to maintain the Linear Correspondence Axiom (LCA). Thiersch discusses and analyzes three languages for which remnant movement has been proposed within the confines of the antisymmetric approach: Malagasy (Rackowski and Travis 2000), Hungarian (Koopman and Szabolcsi 2000), and German (Müller 2004). The main question all these researchers face is "how to create the remnant constituent often enough, and at the appropriate point" (265-66). Embedded in this issue is the difficulty of motivating each step of the movement and the need to create extremely large structures which would allow roll-up as high as is needed. A further complication has to do with the need to induce island effects in some specifier positions but not in others. The resulting system has to accommodate a large number of assumptions. In addition to antisymetry and the LCA, the system must include specific conditions on phases (especially for Müller), particular mechanisms for deriving scrambled orders in Hungarian and German, and ad hoc conditions on extraction which may violate established syntactic principles. Two overall conclusions seem to follow: first, if one is committed to antisymmetry, a principled theory of remnant movement is needed; and second, in the absence of such a theory, existing solutions for deriving verb-initial orders via verbraising may need to be reconsidered. In a way, this latter result gives more weight to the proposals advocating VP-raising (see sections 2.2.4 and 2.2.5 in particular)-although the VP-raising approach is not without flaws either. The order of the chapters in this collection is alphabetical by the name of the author, but my recommendation to an interested reader is to start with Thiersch's chapter because it helps to set the stage for more specific discussions in the remaining parts of the book. This chapter is not the most user-friendly, and it is addressed to an audience deeply entrenched in minimalism, but it presents a useful illustration of how a guiltless little addition to the LCA may unreasonably complicate the resulting theory.
In the chapter "The guest playing host: Adverbial modifiers as matrix verbs in Kavalan," Henry Chang examines inflected adverbs in Kavalan and argues that they are actually verbs; this is meant to explain the appearance of dedicated verbal morphology (e.g., agent focus marking) on Kavalan adverbs. If this approach is on the right track, then what is usually expressed by adjunction in more familiar languages gets expressed via verb serialization or complex predicate formation in Kavalan. In the latter case, the apparent adverb is actually the verbal head, which leads to a reversal of traditional roles: the guest starts playing host. For example, in frequency expressions, matrix focus/aspectual markers and pronominal markers can appear on either "verb" without any change in meaning: 1 (4) a. Pataz-ti-ku s<em>upas tu qRitun.
often <AF>-buff-ASP-1SG.NOM OBL car 'I often buffed the car.' (Chang's ex. [60a, a'], p. 69) Interestingly, Kavalan verbs with adverbial semantics differ in the extent to which they exhibit standard verbal properties. Manner adverbials are morphosyntactically most similar to all other verbs, and higher adverbs ("speech act expressions" per Chang) are less verb-like.
Chang states that Kavalan is typologically unique; in his view, the verbal status of adverb-like expressions entails that they are not relevant for the debate between the feature-based approach and the semantic approach outlined above. However, in the next chapter, "Seediq-adverbial heads in a Formosan language," Arthur Holmer shows that the Atayalic language Seediq also has verbs where one would expect adverbs. Furthermore, Holmer's examples suggest that these adverbial verbs are mostly found in the encoding of manner, although my conclusion here may be just an artifact of his choice of data. Holmer goes on to discuss possible analyses of these data. He presents two options: either adverbial verbs are true verbs and as such undergo verb raising typical of head-initial languages, or they are instantiations of adverbial heads represented in the functional structure à la Cinque (feature-based approach). Holmer chooses the second option, but in doing so he also suggests a revision of the "roll up" structure proposed in Rackowski and Travis (2000) for Malagasy. He proposes to divide syntactic heads in Seediq into two classes, based on the following criteria: (5) X and Y heads in Seediq (after Holmer's (41), p.112) X HEADS Y HEADS Position with respect to the verb Preverbal Postverbal Subject to Head Movement Constraint Yes No Trigger predicate raising No Yes It remains to be seen if evidence for this distinction may be found outside Seediq. At this point, I suspect that an interest in keeping the feature-based approach to adverbs at work in Seediq drives Holmer's distinction.
In the chapter "Patterns of phrasal movement: The Niuean DP," Arsalan Kahnemuyipour and Diane Massam address the structure of the Niuean DP. The structure of Austronesian DPs is understood even less than the structure of Austronesian clauses, so its study is a very important endeavor. The authors subscribe to the notion that the order of elements inside a DP is universal (Greenberg 1966, Cinque 1999, as shown in (6a). They then derive the order of elements in the Niuean DP (6b), a mirror image of the universal order, by successive XP raising: (6) a. Demonstrative > Numeral > Adjective > Noun b. Noun > Adjective > Numeral > Demonstrative The overall pattern of XP movement inside the DP is consistent with the analogous pattern of movement in the verb phrase (structures [14b, c], p. 136), and thus supports the analysis advanced in Rackowski and Travis (2000), with some additional mechanisms required. On such an account, one would expect that nonintersective adjectives should be subject to the same strict linearization constraints as other constituents. I am not aware of the relevant Niuean data, but data from Fijian show that this is not the case. Here we have an unexpected turn on the analysis proposed by Kahnemuyipour and Massam who predict just one order: 2 (7) a. na dau cici totolo e liu DET runner fast former 'a former fast runner' (a runner that was formerly fast) b. na dau cici e liu totolo DET runner former fast 'a former fast runner (a fast person that was formerly a runner)' Every reviewer is tempted to propose an alternate analysis, and this reviewer is no exception. A possible alternative to the roll-up analysis advanced in this chapter is to analyze at least some of the nominal modifiers as reduced relative clauses. If this were the case, the (6b) order in Niuean, where all the modifiers follow the nominal head, would follow and would not require additional assumptions or movement. Austronesian relative clauses are strictly postnominal; complementizers in these clauses delete quite regularly, and their stacking may be subject to simple semantic constraints. With the numerals, their ability to function as predicates (and therefore to be available for relative clause formation) is well-attested-in fact, the authors discuss the sentential predicate function of the numerals on p.141. For adjectives, several researchers have pointed out that distinguishing between true adjectives and reduced relatives in Austronesian is challenging, and it may well be that the distinction is not necessary: just as adverbial meanings may be expressed by verbs in Kavalan and Seediq, maybe the adjectival and numerative meanings are expressed verbally and require a reduced relative.
I am not ready to defend this proposal in detail, but I consider it a virtue of this chapter that it has provoked the reader to consider the issue further. In sum, the chapter raises important questions concerning the structure of DP in Austronesian and will motivate other researchers to look at the DPs in their languages more closely. The attention to flexibility of word order with numerals (section 5 of the chapter), for instance, is particularly commendable; if other Austronesian languages show flexibility of order within DP with some classes of constituents but not others, this will lead us to uncover finer details of language-specific and universal design.
The Seediq chapter and the Niuean chapter take a particular position on the debate between the feature-based and semantic approaches to adverbial syntax. The chapter by Daniel Kaufman, "Rigidity versus relativity in adverbial syntax: Evidence from Tagalog," presents an explicit comparison between the two approaches and uses Tagalog data to argue in favor of the semantic approach. Tagalog scrambling is notoriously difficult to explain, and maintaining the tenets of the antisymmetrical approach in accounting for it creates numerous problems. Kaufman abandons antisymmetry and shows that this is a 2. I would like to thank Dennis Ott for bringing this contrast to my attention.
small price to pay for the correct placement of Tagalog adverbs. Another important contribution of this chapter consists in its attention to the prosodic component of scrambling; in Tagalog, monosyllabic clitics have to precede disyllabic ones, and that requirement immediately creates two separate prosodic domains (Kaufman's [59], p. 176) with which the competing approaches need to contend. The semantic approach outlined by Ernst actually includes a component that is supposed to address ordering preferences based on weight, but it is not always clear how strong these preferences are expected to be, and what predictions could be made for a head-initial language. It seems that a combination of the semantic approach, prosodic factors, and, possibly, the availability of different positions for adverbs (which would suggest that they are all base-generated) could provide a breakthrough for a language like Tagalog. It remains to be seen whether the same approach could be used for more rigidly linearized languages such as Malagasy. Kaufman has a short postscript on Malagasy, but it is too sketchy to determine if the same approach would work.
The chapters by Lisa Travis and by Eric Potsdam address the issue of apparent whmovement in Austronesian, focusing on Malagasy. In the chapter "Voice morphology in Malagasy as clitic left dislocation or through the looking glass: Malagasy in Wonderland," Travis addresses two of the overarching issues identified by the editors: verb-initiality and the behavior of adjuncts with respect to extraction. She sets the ambitious goal of developing a general typology of languages based on their use of features that drive movement, and proposes two types: English-like (her A-type) and Malagasy-like (B-type). In A-type languages, movement targets V heads and DPs; in B-type languages, movement targets D heads and VPs. Further testing of this typology should come from the examination of other languages to determine if the two types proposed are exhaustive.
Travis also makes a point with respect to the structure of focus constructions in Malagasy. Following Paul (2001), and echoing Potsdam's chapter, she shows that these constructions are clefts (more accurately, pseudoclefts) and so do not involve any movement to the sentence initial position. The initial phrase is the predicate and the remaining material is the DP subject, which includes a headless relative clause. The resulting word order is predicate-subject, as is generally found in Malagasy (VS).

(8) [ IP [ PredP DP/PP] [ DP no [ CP …]]]
FOCUS TOPIC Contrary to the traditional voice-based approach to Malagasy verb forms, the morphology on the predicate of a headless relative clause is analyzed not as voice but as agreement with the head of the clause. Travis compares this morphological behavior to cliticization, as in clitic dislocation constructions of more familiar languages, but as I understand it, her proposal embodies the idea that exponents of Malagasy "voice" represent the inflectional morphology resulting from wh-agreement (Chung 1998 and references therein). Thus, her proposal is close to that by Pearson-she cites his UCLA dissertation but not the 2005 paper which spells out the wh-agreement proposal in more detail-who follows the spirit of Chung's proposal concerning wh-agreement in Chamorro. Travis also discusses some finer issues of Malagasy clefts and the interaction between the applicative form (so-called circumstantial voice) and the appearance of certain PPs in sentence-initial position. The difficult question of why some adjuncts are immune to the subject-only constraint while others are not remains unanswered, but the analysis of Malagasy data undertaken in this chapter will no doubt help us all look into finer details of adjunction in Austronesian.
Eric Potsdam's chapter, "The cleft structure of Malagasy wh-questions," portrays Malagasy as a true wh-in-situ language. Potsdam shows that wh-questions in Malagasy are not derived by A-bar movement, but instead constitute pseudoclefts with the basegenerated wh-phrase in the predicate position and a DP containing a headless relative clause in the subject position. Compare the focus structure in (8)  The arguments for this analysis are very data-rich, and Potsdam develops a number of useful diagnostics that could easily be applied to other languages. In addition to presenting a compelling analysis of single wh-questions, he addresses the issue of supposed multiple wh-questions in Malagasy (Sabel 2003), as in the following: (10) Aiza iza no mividy ny vary?
where who PRT buy DET rice 'Who buys rice where?' (Potsdam's ex. [49], p. 216) Assuming that something like (10) is acceptable (in my experience, many speakers cringe hearing this), Sabel uses such data to argue for optional multiple wh-fronting, a rather unexpected turn of events for an Austronesian language. Potsdam's analysis of these questions maintains the structure in (9) and simply places the second wh-word inside the subject part of the question-it is thus similar to a possessor, and the whole question in (10) could be literally translated as "Where is whose buying of rice?". The arguments for this approach include substitution with non-wh-expressions, coordination, and particle placement. Potsdam then goes on to show that his analysis does not face the same theoretical difficulties as the multiple wh-fronting analysis and does not require any additional stipulations.
The outstanding problem for the analyses proposed by Travis and Potsdam is to determine the category of the particle no that appears in the examples above. The possibilities include either a determiner head or a complementizer, but at this juncture it is hard to tell which of the two analyses fares better. If we view no as a determiner, the difficulty lies in explaining why it is found exclusively in the subjects of pseudoclefts. If it is a complementizer, its distribution is far from straightforward.
To recapitulate, recall that the charge from the editors to the authors was to consider the licensing of adjuncts and to account for the ability of adjuncts to escape the subjectonly constraint. How did the chapters in the book fare with respect to these two main issues? Chang, Holmer, Kahnemuyipour and Massam, Kaufman, Thiersch, and to some degree Travis all touch upon the issue of licensing and linearization. Except for Kaufman, most chapters just adopt one of the two main approaches to adverbial syntax and follow their theory of choice. The papers by Potsdam and Travis bear on the second issue, that of adjuncts and the subject-only restriction, and although they do not resolve it, they offer the important conclusion that one does not have to posit an unbounded dependency lightly-sometimes, what may seem to be an instance of an unwieldy A-bar movement turns out to be a compliant base-generated structure.
Taken as a whole, this book represents an advance in the theoretical discussion of the clause structure, licensing, and distribution of adjuncts. The empirical data are rich and often new, and bringing them to the attention of linguists who may not otherwise peruse individual grammars is a major step forward. More importantly still, the collection introduces a number of interesting proposals which are likely to serve as the point of departure for further discussions. For example, Kalin (2009) builds her analysis of Malagasy pseudoclefts on Potsdam's work. Most chapters in the book are easy to follow, and the Austronesian data are presented in a manner that should make them accessible to those who have never seen a head-initial language with infixation and intricate voice morphology.
It is inevitable that a book like this cannot include all the necessary information, but there are several aspects that I found particularly wanting. First, the book could have benefited from more discussion of Oceanic languages, especially since the literature on a number of these languages is quite impressive. Without such breadth, a reader may get the impression that all the action in theoretical work happens in Western Malayo-Polynesian, which is certainly not the case. Second, the book could have done more to bring together minimalist explorations in Austronesian with work done in other frameworks-for example, LFG and functionalist approaches. Some examples of the former include the work on various Austronesian languages by Peter Austin, Paul Kroeger, Simon Musgrave, Wayan Arka, and Peter Sells. The chapter by Chang relates the Kavalan material to functionalist explanations in terms of clausal connectedness, but as such it stands alone in the volume. Meanwhile, several functionalist researchers have done important work on Austronesian (the introductory chapter refers to some of that work, in particular to the work of Nikolaus Himmelmann). The approaches and the solutions may differ, but the questions address similar themes: the nature of verb-initiality, possible accounts of the subject-only restriction, and the distribution of determiners. If anything, generative and functionalist approaches complement each other in the way they approach data collection. Functionalists rely on naturally occurring data and often find controlled elicitations unnatural, whereas more formal approaches emphasize controlled elicitations. Combining the two approaches may be beneficial, if only in informing us whether certain patterns are categorical or gradient. If categorical, a grammatical explanation is all the more likely. For instance, if a language allows only VOS but not VSO word order, this may provide an additional argument in favor of a particular way of deriving the verb initial order; see options (ii)-(v) above.
Given that most of the chapters focus on the tension between the feature-based and semantic accounts of adverbs, it would have been productive to have the papers crossreference each other more, and engage in a dialogue not only with established proposals in the field but with each other. (Thiersch refers to Kaufman's paper, but that is more of an exception. Holmer refers to a paper by Chang which seems like a precursor to the latter's paper in this volume.) Without articulation across the chapters, the book reads more as a set of proceedings (which to some extent it is, as it is based on several papers presented at a conference in Berlin in 2004).