Long-Distance Dependencies in Tagalog: The Case for Raising Veronica Gerassimova and Peter Sells Stanford University and SOAS, University of London 1. Introduction One famous feature of Tagalog and other Western Austronesian languages is that only the Subject1 may relativize (e.g., Schachter & Otanes 1972); see (1-3). The verb's voice picks out one argument as Subject, which is the missing constituent in a relative clause; hence (3) is ungrammatical:2 1. Binigyan ng sundalo ng pera ang babae. gave.RV GEN soldier GEN money NOM woman The soldier gave the money to the woman. 2. ang babae =ng binigyan ng sundalo ng pera NOM woman =L gave.RV GEN soldier GEN money the woman that the soldier gave the money to 3. *ang babae =ng ibinigay ng sundalo ang pera NOM woman =L gave.PV GEN soldier NOM money Intended: the woman that the soldier gave the money to Subject: Recipient Subject: Recipient Subject: Theme Another notorious characteristic of Tagalog relativization is the requirement that the predicates in a long-distance chain each appear in a particular voice (usually Patient voice), and that none of their overt arguments be a Nominative DP (Sells 2000). We refer to such chains as Subject-only chains. Now Subject-only chains also uncontroversially characterize raising constructions in Tagalog (Davies 2000, Nakamura 2000, Sells 2000). In order to account for the identical constraints on relativization and raising, some have proposed that raising is A-bar-movement (e.g., Nakamura 2000). Following Gerassimova (2005), we explore the opposite possibility – that all Subject-only chains are in fact raising dependencies. This not only solves the problems faced by A-bar movement analyses, but also has desirable consequences for several larger issues, including the constraints on whmovement in Universal Grammar, the consequences that one can draw from the Keenan-Comrie Accessibility Hierarchy, and the claimed link between extraction and voice in so-called wh-agreement phenomena. While the Tagalog system is assumed to be essentially different from what we find in English, English does in fact show voice constraints in relativization, specifically, in so-called reduced relatives: 4. a. b. c. the man __ known by Mary __ to be studying French the man __ known by Mary __ to have been photographed by Sam the man __ known by Mary *(that Sam photographed __ ) English reduced RCs The contrast in the b/c examples is rather similar to what we find in Tagalog; the apparent voice constraints are not surprising in English, for we would immediately diagnose these dependencies as involving raising, not extraction. These examples provide a direct model for our approach to Tagalog. As well as arguing that there is no A-bar movement in Tagalog relative clauses, we show that whphrases in questions appear in only and exactly the same places where their non-wh-counterparts do (see also Gerassimova 2005). Hence Tagalog is a wh-in-situ language, and thus more similar to 1 The argument marked with Nominative ang (si for names), also known as focus, topic, absolutive or trigger. 2 Glosses: AV – Actor voice; DAT – Dative case; DV – Dative voice; GEN – Genitive case; L – linker; NASA – locative predication form; NOM – Nominative case; PV – Patient voice; PL – plural; Q – question particle; RV – Recipient voice; [] – clause boundary; = – cliticization site; S&O – Schachter and Otanes. © 2008 Veronica Gerassimova and Peter Sells. Proceedings of the 26th West Coast Conference on Formal Linguistics, ed. Charles B. Chang and Hannah J. Haynie, 190-198. Somerville, MA: Cascadilla Proceedings Project. 191 languages like Malagasy (Sabel 2003, Paul & Potsdam, to appear), Niuean (Massam 2002), and Indonesian (Saddy 1991; Cole at al. 2005) than previously thought. Simply put, we argue that Tagalog has no evidence of any A-bar movement and (correspondingly) no evidence of overt wh-movement. Section 2 concerns relativization in Tagalog, with a review of existing extraction analyses of Subject-only chains; we then present our own analysis of Subject-only chains as raising chains. In Section 3 we turn to various strategies for wh-question formation, none involving movement, including the fronting of wh-adverbs and in-situ argument wh-phrases. Finally, in Section 4 we consider the bigger picture of (presumed) extraction in Tagalog, and discuss the consequences of our analysis for Universal Grammar, the Keenan-Comrie Hierarchy, and the typology of wh-agreement phenomena. 2. Tagalog relativization: Subject-only chains as raising In Tagalog, verbs in Subject-only chains are required to appear in a particular voice form. In the immediate clause containing the gap, or the base clause, this is the voice form that identifies the empty position as the Subject of the clause, as in the relative clause examples in (1-3) above. In higher clauses along the chain, verbs must appear in one particular voice, usually the Patient voice:3 4 5. ang kalabaw [na ipinangako ng guro [na bibigyan ng lalaki ng bulaklak]] NOM water.buffalo L promised.PV GEN teacher L gave.DV GEN man GEN flower the water buffalo [(that) the teacher promised [(that) the man would give a flower to]] 6. ang babae [na sinabi ko kay Pedro [na ikakasal kay Jose]] NOM woman L said.PV I.GEN DAT Pedro L marry.PV DAT Jose the woman [that I said to Pedro [that will marry Jose]] 2.1 Previous analyses: extraction In much current work, Subject-only chains are treated as extraction, with various extra constraints to account for the fact that only Subjects may extract and for the voice constraints along the chain (e.g., Kroeger 1993, Nakamura 2000, Richards 2000, Aldridge 2002, Paul 2002, Rackowski 2002, Rackowski & Richards 2005). According to extraction analyses, Subject-only chains are formed via wh-movement of a null Operator; the voice of the base clause identifies as its Subject the empty position; and the voice of a predicate in a higher clause identifies as its Subject its complement clause, through which the dependency passes. This style of analysis necessitates peculiar constraints on A-bar dependencies: that only Subjects may extract, and extraction is allowed only out of sentential Subjects, resulting in the following constraints on A-bar dependencies: 7. a. A gap must always be the Subject in its immediate clause, the base clause. b. A subordinate clause containing a gap must itself be the Subject of its containing clause. Such extraction accounts raise both theoretical and empirical problems. First, they imply that an A-bar dependency interacts with voice. While voice is expected to play a central role in Adependencies, it is not in A-bar extraction – this is one of the empirical motivations for the theoretical distinction between A- and A-bar dependencies. Thus the claim that the voice of predicates in Subjectonly chains is determined by extraction obliterates a core difference between A- and A-bar constructions. Moreover, treating Subject-only chains as extraction results in a violation of the familiar constraint on extraction from sentential subjects, attested in a number of languages. Although there are languages in which sentential subjects are not extraction islands, extraction analysis of Tagalog Subject-only chains is unique in that sentential subjects are the only possible environment for extraction, and it “not only allows, but actually requires, violations of the Sentential Subject 3 In both raising and relativization, clauses in the chain are introduced by the linker/complementizer na/-ng. 4 Any other voice form on the verbs in (5-6) leads to ungrammaticality. These restrictions on voice are only found in dependency constructions; regular subordinate clauses allow any voice form. 192 Constraint” (Kroeger 1993:210). Another undesirable consequence of extraction analyses is that Subjects are commonly dispreferred as extractees across languages. In a number of languages, objects are easier to extract than Subjects. In the English that-trace phenomenon, for instance, when an overt complementizer is present, subjects (8) but not objects (9) are precluded from extracting. Tagalog, on the other hand, is ascribed the unusual typological property of disallowing object extraction and requiring Subject extraction. 8. *Who do you believe that __ saw Jim? 9. Who do you believe that Tom saw __? Finally, extraction accounts fail to explain a key fact about the distribution of Subject-only chains, noted by several authors (Davies 2000, Nakamura 2000, Sells 2000) – in some languages, including Tagalog and Cebuano, chains with voice constraints occur in raising as well as relative clauses; some researchers have concluded that both relative clauses and raising are instances of A-bar extraction. To deal with these issues, we propose an alternative account of relativization in Tagalog. We suggest that the Subject-only chains found in relative clauses in fact involve (Subject-to-Subject) raising. A raising account of Subject-only chains avoids all of the undesirable effects of the extraction analyses, and provides a straightforward explanation of the otherwise puzzling properties just noted. 2.2 Our proposal: Subject-only chains as raising As already mentioned, Subject-only chains have the same structure as raising chains (Davies 2000, Nakamura 2000, Sells 2000) – in both constructions it is the Subject of the base clause that is targeted by the dependency, and verbs higher up in the chain must appear in a particular voice form. Raising in Tagalog is always Subject-to-Subject. In (10) all of the arguments of the embedded clause are in their normal position. In (11) the Nominative of the complement clause, ang panbansang awit ‘national anthem’, has been raised into the matrix clause, and the downstairs Nominative position is empty. In contrast, (12) is ungrammatical because the raised phrase, Linda, is not the downstairs Subject. 10. Inasahan ko [na awitin ni Linda ang pambansang awit]. expected.RV I.GEN L sang.PV GEN Linda NOM national anthem I expect (for) Linda to sing the national anthem. 11. Inasahan ko ang pambansang awit [na awitin ni Linda]. expected.RV I.GEN NOM national anthem L sang.PV GEN Linda I expected the national anthem to be sung by Linda. (Kroeger 1993:28, ex. 17) 12. *Inasahan ko si Linda [na awitin ang panbansang awit]. expected.RV I.GEN NOM Linda L sang.PV NOM national anthem Intended: I expected Linda to sing the national anthem. (Kroeger 1993:28, ex. 17) And like relativization, shown in (6) (repeated below in (13)), raising may also take place across more than one clause. In (14) the DP ang babae ‘the woman’, which is semantically an argument of the lowest clause, is raised two clauses up and appears as an element of the main clause: 13. ang babae [na sinabi ko kay Pedro [na ikakasal kay Jose]] NOM woman L said.PV I.GEN DAT Pedro L marry.PV DAT Jose the woman [that I said to Pedro [that will marry]] 14. Pinaniwalaan ni Maria ang babae [=ng sinabi ni Juan [na binigyan ng pera]]. believed.RV GEN Maria NOM woman L said.PV GEN Juan L gave.RV GEN money Maria believed the woman to have been said by Juan to have been given (the) money by him. One unappreciated fact about Subject-only chains is that they display not only the form, but also the distribution of raising dependencies, a property also found in English reduced relatives (see (4) above). Considering the voice forms of verbs in English Subject-to-Subject raising, such as the ones in (4) above, it is clear that they obey the same constraints as verbs in Tagalog: in the base clause the raised phrase is always the Subject, while in higher clauses the verbs are always in the passive voice, having as their Subject an empty argument position, as in (16), which is the structure of (4c). 193 15. [ __SUBJ known by Mary [ __ SUBJ to have been photographed by Sam]] Hence we propose that higher clauses of Subject-only chains contain an empty Subject position, which is the subject of verbs in Subject-only chains, as in typical raising (Gerassimova 2005): 16. [na L sinabi ko kay Pedro said.PV I.GEN DAT Pedro __ SUBJ [na L ikakasal marry.PV kay Jose __ SUBJ]] DAT Jose The generalization covering all dependencies in Tagalog is that there can be no overt Nominative DP within any of the clauses into which the dependency passes (Sells 2000), and the apparent voice constraints always conspire to make this so. To see how the voice constraints work, imagine that sinabi “say.PV” in (16) were in its Actor voice form, namely, nagsabi; in that case Pedro would be the Nominative Subject of the higher clause, and the dependency chain would not be possible: ikakasal 17. *[na nagsabi ko ang Pedro __ SUBJ [na L said.AV I.GEN NOM Pedro L marry.PV Intended: [said by me to Pedro __ [to have been given the money]] kay Jose __ SUBJ]] DAT Jose The voice form of each intermediate verb must be such that none of the overt DP arguments is picked out as Subject – on our account, it is the missing constituent which corresponds to the Subject. In sum, our account of Subject-only chains as raising explains the voice and Subject restrictions – puzzling and typologically odd under an extraction account – as natural epiphenomena of standard raising, and shows that they in fact constitute a familiar and well-studied type of syntactic dependency. 3. Wh-questions: the case against movement Subject wh-questions are formed as pseudo-clefts, consisting of two juxtaposed DPs – an initial predicate followed by a nominalized relative clause (Paul 2001, Potsdam 2004). Hence the familiar voice constraints hold inside the relative clause part, as in these examples of a wh-cleft: 18. Ano ang binili what NOM bought.PV What did Maria buy? ni Maria? 19. *Ano ang bumili si Maria? GEN Maria what NOM bought.AV NOM Maria Intended: What did Maria buy? The focused element ano 'what' is the matrix (copular) predicate, and the clause final Subject is a headless relative clause, with the case marker ang at its left edge. There is fairly clear consensus among Austronesian researchers that the wh-phrase in these examples is generated in situ. In contrast, non-Subjects form questions via several different strategies, probably none of which has been fully systematically explored; tellingly, voice constraints do not apply in any of these cases. 3.1 Focus Fronting One type of question with a sentence-initial wh-phrase involves Focus Fronting, also known as Emphatic Inversion (Schachter & Otanes 1972:496-8), Adjunct Fronting (Kroeger 1993:43-8), or PPfocus (Mercado 2004). An adjunct or oblique appears in pre-verbal position, receiving an information focus interpretation (e.g., Schachter & Otanes 1972, Mercado 2004), as in (21): 20. Binili ni Maria ang laruan bought.PV GEN Maria NOM toy Maria bought the toy for Pedro. 21. Para kay Pedro binili ni Maria for DAT Pedro bought.PV GEN Maria It's for Pedro that Maria bought the toy. para for kay Pedro. DAT Pedro ang laruan. NOM toy Several tests indicate that the Focus position is not clause-external, but is a part of the minimal 194 clause. One type of evidence comes from the placement of second-position clitics, which are always placed with respect to IP, following its first constituent (see Kroeger 1993). In Focus Fronting, in which a preposed phrase appears immediately before the predicate, second-position clitics appear after right after that fronted phrase (the clitics are underlined in the examples below): mangangisda? 23. Saan ka pupunta? 22. Kailan siya when he.NOM fish.AV where you.NOM go.AV When will he go fishing? (S&O 1972:505) Where are you going? (S&O 1972:514) Clitic placement diagnoses the focused phrase as in Spec,IP (Kroeger 1993), rather than Spec,CP. Additionally, the Focus phrase always occurs to the right of other types of preverbal phrases, such as those found with ay-inversion (25) and topicalization (Kroeger 1993): 24. Kakain kami eat.AV we.NOM We'll eat here. dito. here.DAT 25. Kami='y dito kakain. we.NOM=AY here.DAT eat.AV It's here that we'll eat. (S&O 1972:498) The focus position in (21) above is a regular clause-internal position for adjuncts of various kinds, including all the relevant wh-phrases. Given the interpretive connection between wh and Focus, it is natural that wh-phrases will be in this position. For our argument, the key observation is that adjunct wh-phrases in (22-23) have the same distribution as their non-wh counterparts. Note that voice constraints play no role in these adjunct fronting constructions. 3.2 Wh-expressions as syntactic predicates Wh-phrases may also appear directly as syntactic predicates, with the structural properties of their non-wh counterparts. They may serve as both non-verbal and verbal predicates. Non-verbal whexpressions, such as adverb, DP and PP wh-phrases, are 1-place predicates. In (26-27) the wh-phrases are in the canonical predicate position: 26. Kailan ang miting? when NOM meeting When is the meeting? (S&O 1972: 505) 27. Tungkol sa ano ang kuwento? about DAT what NOM story What's the story about? (S&O 1972: 507) Direct evidence for the predicate status of the wh-phrases in these examples comes from a special feature of locative phrases – they appear with the case marker sa when they occur as arguments, as in (28-29), yet with a distinct marker, nasa when they occur as predicates, as in (30-31): 28. Kumakain sila IM.eat.AV he.NOM He eats at the school. sa iskwela. DAT school 30. Nasa Maynila ang gusali. NASA Manila NOM building The building is in Manila. 29. May trabaho siya sa kanila. exist work he.NOM DAT their He has a job at their place. (S&O 1972:450) 31. Nasa opisina si Esting. NASA office NOM Esting Esting is at/in the office. (S&O 1972:256) Crucially, in questions locative wh-phrases also appear with the predicate version of locative marking, as in the following examples: 32. Nasa ano na ba ang pusa? 33. Nasaan si Pedro? (nasaan = nasa + saan) NASA what now Q NOM cat where NOM Pedro On what is the cat now? (S&O 1972:507) Where is Pedro? (S&O 1972:513) In (33) the wh-word ano ‘what’ has the predicate marker nasa, and in (34) the wh-phrase ‘where’, whose basic form is saan, lexically incorporates the marker nasa in the predicate form nasaan. This predicate use contrasts with wh-locatives in questions with verbal predicates; there, whlocatives appear in the simple non-predicative form, saan, via Focus Fronting (discussed above). 195 34. Saan ka pupunta? where you.NOM go.AV Where are you going? (S&O 1972:514) In addition, wh-words in Tagalog may actually serve as verbal predicates. For example, ano ‘what’ can occur as a “substitute for a verb base”: it appears with all verbal morphology, including voice marking, and in the position of verbal predicates (Schachter & Otanes 1972:509), as in (35-36), which are perfectly canonical Tagalog clauses, whose predicate is a wh-predicate. All three examples involve a version of the wh-word ano ‘what’ in a predicative use: it occupies the canonical predicate position at the left edge of the clause, and is marked for the verbal categories voice and aspect: 35. Nagano ka ba? what.AV you.NOM Q What did you do? 36. Inano mo ba ito? what.PV you.GEN Q this.NOM What did you do with this? (S&O 1972: 509) These data again show that wh-phrases occupy positions already available to constituents of the same type, rather than appearing in a special position associated with extraction. 3.3 Wh-expressions as linked modifiers In Tagalog, certain adjuncts, such as time and manner adverbs, do not participate in the fronting constructions discussed so far. Instead they may occur immediately adjacent to the verb as modifiers. In such cases, the adverb may either precede or follow the verb, and it is linked with a na/-ng linker as noted by S&O (1972:437). Consider these examples with the adverb biglan ‘suddenly’: 37. Biglan =ng binuksan ni Fred ang pintuan. suddenly L opened.PV GEN Fred NOM door Fred opened the door suddenly. 38. Binuksan =ng biglan ni Fred ang pintuan. opened.PV L suddenly GEN Fred NOM door Fred opened the door suddenly. (S&O 1972:451) The corresponding class of wh-phrases appear as modifiers as well, preceding the verb, with an optional linker. (The word raw in (39) is a second-position clitic.) 39. Paano raw (na) nakarating ang mga Cruz? 40. Paano siya (=ng) magsayaw? how they.say (L) arrive.AV NOM PL Cruz how he.NOM (L) danced.AV How did they say the Cruzes came? How does he dance? (S&O 1972:514) This observation provides yet one more piece of evidence that wh-phrases in Tagalog appear in the positions that are independently available to their non-interrogative equivalents. 3.4 Wh-phrases in situ Schachter & Otanes (1972), perhaps the most detailed description of wh-questions, only list examples where wh-phrases appear at the left periphery. For arguments, it is assumed that the only way to form a question is to turn the argument into a Subject and cleft it, as in (18) above. Yet recent work shows that there is an alternative – non-Subject arguments may also remain in situ, although more research is needed to establish if and what constraints exist on in-situ core-argument questions: 41. Kumain ng ano si Mark? ate.AV GEN what NOM Mark What did Mark eat? (Dery 2006, ex. 12) 42. Nakita ka nino? saw.PV you.NOM who Who saw you? Interestingly, examples of Subject wh-phrases in situ have also been cited in recent work (Law & Gärtner 2005, Mayr 2006), as in (43-44): 43. Bumili ng isda ang ano? bought.AV GEN fish NOM what Who bought fish? (Law & Gärtner 2005, ex. 15) 44. Binasa ni Jeff ang ano? read.PV GEN Jeff NOM what What did Jeff read? (Mayr 2006, ex. 3) 196 On the assumption that at least some examples of this form are acceptable as true questions, they show that Tagalog has a wh-in-situ strategy for every type of wh-phrase. Our account of wh-questions essentially leaves every independent position available to wh-phrases, similar to the account of whquestion formation in Indonesian proposed by Cole at al. (to appear), who reach the conclusion “the syntax in Indonesian must parallel information structure”. On the interpretive side, we might assume that the surface position of wh has to be a position where Focus can be properly interpreted. Focus positions in Tagalog are typically at the left periphery: predicates, Focus Fronting and clefting all provide the right syntactic conditions (e.g., Mercado 2004, Dery 2007, Nagaya 2007). These are also the preferred positions for wh-phrases, but are not obligatory, as we have seen. There may also be a clause-typing function fulfilled by a wh-phrase in the initial position. On the other hand, Subjects in Tagalog have a relatively close connection to topic interpretation (e.g., Schachter 1976), disfavoring wh-phrases as Subject. (In the canonical cleft wh-questions, the initial wh is a predicate, not derivationally related to the Nominative ‘gap’ inside the following free relative.) In the examples above with Nominative wh-phrases, these phrases have to be given a strong Focus interpretation by non-positional means. And we assume that a focus position is possible inside the clause (with associated accent) (Dery 2007), allowing non-initial wh-in situ. In sum, in this section we have shown, on the one hand, that the left-periphery position of whphrases is licensed by a host of independently attested constructions, including Focus Fronting, linked modification, and syntactic predication, not by extraction, and on the other that wh-arguments, including Subjects, may stay in situ. In other words, wh-phrases always appear in situ. In this respect, Tagalog is similar to other Western Austronesian languages, such as Malagasy, Niuean, Indonesian, Seediq and Madurese. Malagasy (Sabel 2003, Paul & Potsdam, to appear), Niuean (Massam 2002), and Indonesian (Saddy 1991, Cole at al. 2005; to appear). 4. Extraction in Tagalog: the bigger picture We suggest that Tagalog is best described as a wh-in situ language, without even the null operator kind of wh-movement: relative clauses involve only raising chains.5 A non-extraction analysis of Subject-only chains has desirable consequences for what UG is understood to allow. If Tagalog is indeed a wh-in-situ language and no wh-expressions move, an extraction analysis of Subject-only chains (e.g., Aldridge 2002, Rackowski & Richards 2005) means that wh-movement in Tagalog is only available to null elements – in relative clauses. This raises the question of whether we should really expect to find a language such that it allows syntactic whmovement, and yet overt wh-expressions in that language never move. We believe not: we would like to speculate that if a language has A-bar extraction as an option, it should be available as a strategy for wh-question formation as well, i.e., there should be overt evidence for it. Our proposal to treat Subject-only chains as raising, rather than extraction, provides a more coherent perspective on Tagalog. Rather than saying that it has an unusual mixture of whmovement and base-generation strategies, with some mysterious restriction precluding all overt whelements from moving, we can simply state that wh-movement (A-bar-movement) is not available as an option in this language, and possibly in other Western Austronesian languages as well. In fact, in the AGREE-based approach to licensing in current Minimalist syntax, movement itself is triggered by an EPP or OCC feature; movement is now essentially theoretically motivated only when overtly visible. A very surprising fact on the extraction account is that the putative voice constraints on movement hold for argument wh-phrases but not for adjuncts. Some authors, such as Rackowski & Richards (2005), have tried to attribute this contrast between arguments and adjuncts to a distinction between DPs and non-DPs. On that account, the set of DPs would be restricted to Nominatives and Genitives, whereas non-DPs would include the remaining arguments and adjuncts – Datives, PPs, adverbials, etc. Crucially, the contrast between arguments and adjuncts cannot be reduced to DP vs. non-DP, for Dative phrases are clearly DPs, not PPs, yet they pattern as adjuncts (Gerassimova 2005). All formal features of Datives in Tagalog indicate that they are DPs, just like Nominatives and Genitives: First, like Nominative and Genitive case markers, Dative markers are different for common (sa) and proper (kay) nouns. Such differentiation characterizes crosslinguistically case markers, but not prepositions; 5 One may wish to posit a null operator in a relative clause for the purposes of semantic interpretation. This would not affect our argument, as we only claim that there is no evidence for movement of a null operator. 197 Tagalog prepositions do not display such alternations either. Second, phrases marked with sa may appear in the same environments as other DPs, e.g., either a Dative or a Genitive marked-phrase may occur in the comparative construction. Another indicator that Dative phrases are DPs is that canonical prepositions take as their complements Dative phrases, e.g., para sa N “for N”, buhat sa N “from N”, galing sa N “from N”, mula sa N “from N”, “about N”, etc. (Schachter & Otanes 1972:260). Finally, Dative sa displays a wide variety of functions – Locative, Benefactive and Goal arguments, as well as definite non-Nominative patients, which is typical of case markers but uncharacteristic of prepositions. Thus the fact that voice constraints hold for arguments but not adjuncts does not correlate with DP vs. non-DP, and hence cannot be related to Case or some other kind of ‘agreement’-type licensing. The mysteries inherent in the extraction approach all disappear if arguments may be questioned literally in-situ, or via the preferred cleft construction, in which a wh-phrase is base-generated initially. Adjunct questions involve a variety of (different) in-situ strategies, as we showed above. In their influential paper, Keenan & Comrie (1977) suggest that each language specifies the syntactic roles that may relativize, determined by a span of the Accessibility Hierarchy (AH), in (45): 45. Subject > Direct Object > Indirect Object > Oblique > Genitive > Object of Comparison If a relativization strategy applies to a particular position on the AH, it will also apply to all of the higher positions. The majority of the languages in Keenan and Comrie’s sample which have ‘Subjectonly’ for the primary relativization strategy are Austronesian. From the perspective we take here, the ‘Subject-only’ constraint sheds no light on relativization or on the typology of relativization strategies: it is a consequence of the fact that the Philippine-type voice system generates many kinds of one-place predicates – where a ‘predicate’ and a Nominative Subject comprise a clause. The key observation – made by Keenan himself in later work (Keenan 1995, 2005) – is that such predicates can be directly used or chained in raising and relative clauses. Hence, there are no special restrictions on extraction limiting it to Subjects, but rather these languages have flexible ways of creating predicates, within the voice system and without extraction. Consequently, the constraints on relativization in Tagalog follow directly from the constraints on voice, and so any reference to the Accessibility Hierarchy is unnecessary, and misleading. Further, concerns about how to enforce at a theoretical level extraction of Subjects but not objects, or extraction out of sentential Subjects but not sentential complements simply become irrelevant. One important conclusion from our proposal is that voice does not, after all, interact with extraction, something which has only been claimed for Western Austronesian languages anyway. This allows us to maintain the position that A- and A-bar dependencies are fundamentally different, and that argument realization patterns are relevant only for the former. Although we have focused on Subject-only chains in Tagalog, the raising account of relativization can be naturally extended to the very similar dependencies reported for a variety of other languages, most notably the Philippine languages, including Ilocano, and Cebuano; the Javanic languages, including Malay, Javanese, Madurese, Balinese, and Indonesian; and Malagasy (e.g., Davies 2000, Davies and Dubinsky 2004). Furthermore, our analysis of Subject-only chains as raising provides important insights into the apparent link between extraction and voice in the famous so-called wh-agreement phenomenon assumed for Chamorro (Chung 1982, 1998, Chung and Georgopoulos 1988, Georgopoulos 1985, 1991) and Palauan (Chung and Georgopoulos 1988, Georgopoulos 1985, 1991), which were also the original inspiration for the extraction analysis of Subject-only chains. According to the wh-agreement analysis, there are verb forms that are only used in extraction dependencies, and whose sole function is to indicate the syntactic role of the gap. The absence of extraction in Tagalog is consistent with recent findings about the absence of wh-agreement in Chamorro and Palauan, and suggests that whagreement phenomena involving voice are the result of misinterpreting raising dependencies in all of these languages as extraction. Donohue and Maclachlan (1999) and Gerassimova (2002, 2005) demonstrate that in both Chamorro and Palauan, respectively, the assumed wh-agreement verb forms in fact mark Philippine-type voice.6 In her analysis of Palauan, Gerassimova (2005) argues that Palauan relativization is best analyzed along the lines of Subject-only dependencies found in Tagalog 6 The gist of both accounts is that so-called wh-agreement verb forms are neither necessary nor sufficient markers of filler-gap dependencies. Rather they are remnants of an older, Philippine-type voice system, and are today extant in (a variety of) subordinate clauses, including relative clauses. The illusion that these forms are unique to filler-gap dependencies is due to the inability of these older forms to appear in matrix clauses. 198 relativization, and that the apparent wh-agreement forms are in fact voice forms constraining the relative prominence of Subject and non-Subject dependents. References Aldridge, Edith. 2002. Nominalization and WH-movement in Seediq and Tagalog. Language and Linguistics 3, 393-426. Chung, Sandra. 1982. Unbounded dependencies in Chamorro grammar. Linguistic Inquiry 13, 39-78. Chung, Sandra. 1998. The Design of Agreement: Evidence from Chamorro. Chicago, U. of Chicago Press. Chung, Sandra and Carol Georgopoulos. 1988. Agreement with gaps in Chamorro and Palauan. In M. Barlow and C. A. Ferguson (eds.) Agreement in Natural Language: Approaches, Theories, and Descriptions. Stanford, CSLI Publications, 251-267. Cole, Peter, Gabriella Hermon and Yassir Tjung. 2005. How Irregular is WH in Situ in Indonesian? Studies in Language 29, 553-581. Cole, Peter, Gabriella Hermon and Norhaida Bt. Aman. to appear. Clefted questions in Malay. In David Gil (ed.) Malay/Indonesian Linguistics. London, Curzon Press. Davies, William. 2000. Against raising in Madurese (and other Javanic languages). CLS 36. Davies, William and Stanley Dubinsky. 2004. The Grammar of Raising and Control. Oxford, Blackwell. Dery, Jeruen. 2007. Focus and Constituent Order in Tagalog. International Conference on Role and Reference Grammar, Institute of Linguistics, Academia Sinica, Taipei, Taiwan ROC. Donohue, Mark and Anna MacLachlan. 1999. What agreement in Chamorro? In Toronto Working Papers in Linguistics, vol. 16.2. Proceedings of AFLA VI, 121-132. Georgopoulos, Carol. 1985. Variables in Palauan Syntax. Natural Language and Linguistic Theory 3, 59-94. Georgopoulos, Carol. 1991. Syntactic Variables: Resumptive Pronouns and A’-Binding in Palauan. Dordrecht, Kluwer Academic Publishers. Gerassimova, Veronica. 2005. Unbounded Dependency Constructions in Western Austronesian. Ph.D. Thesis, Stanford University. Keenan, Edward and Bernard Comrie. 1977. Noun Phrase Accessibility and Universal Grammar. Linguistic Inquiry 8, 63-99. Keenan, Edward. 1995. Predicate-argument structure in Malagasy. In C. S. Burgess (ed.) Grammatical Relations: Theoretical Approaches to Empirical Questions. Stanford, CSLI Publications. Kroeger, Paul. 1993. Phrase Structure and Grammatical Relations in Tagalog. Stanford, CSLI Publications. Law, Paul and Hans-Martin Gärtner. 2005. UCLA Working Papers in Linguistics, no. 12, Proceedings of AFLA XII, Heinz and Ntelitheos (eds.), 211-226. Massam, Diane. 2002. Troublesome Questions in Niuean. Paper presented at AFLA 9. Mayr, Clemens. 2006. Extraction in Tagalog, ms. Harvard University. Mercado, Rafael. 2004. Focus constructions and WH-questions in Tagalog: A unified analysis. Toronto Working Papers in Linguistics 23.1, 95-118. Nakamura, Masanori. 2000. On raising in Tagalog. Lingua 110, 391–408. Nagaya, Naonori. 2007. Information structure and constituent order in Tagalog. Lang. and Linguistics 8, 343-372. Paul, Ileana. 2001. Concealed pseudo-clefts. Lingua 111, 707–727. Paul, Ileana. 2002. On extraction asymmetries. In MIT WPL 44: The Proceedings of AFLA XIII, 211-224. Paul, Ileana and Eric Potsdam. to appear. Sluicing without Wh-Movement in Malagasy. In Jason Merchant (ed.) Sluicing: Cross-Linguistic Perspectives. Amsterdam/Philadelphia, John Benjamins. Potsdam, Eric. 2004. Wh-questions in Malagasy. In Paul Law (ed.) Proceedings of AFLA XI. ZAS Papers in Linguistics 34/10. Berlin: Zentrum für Allgemeine Sprachwissenschaft, 244-258. Rackowski, Andrea. 2002. The Structure of Tagalog: Specificity, Voice, and the Distribution of Arguments. Ph.D. Thesis, MIT. Rackowski, Andrea and Norvin Richards. 2005. Phase Edge and Extraction: A Tagalog Case Study. Linguistic Inquiry 36, 565-599. Richards, Norvin. 2000. Another Look at Tagalog Subjects. In Ileana Paul, Vivianne Phillips and Lisa Travis (eds.) Formal Issues in Austronesian Linguistics. Dordrecht, Kluwer Academic Press, 105-116. Sabel, Joachim. 2003. Malagasy as an optional multiple wh-fronting language. In Cedric Boeckx and Kleanthes Grohmann (eds.) Multiple wh-fronting. Amsterdam/Philadelphia, John Benjamins, 229-254. Saddy, Douglas. 1991. WH Scope Mechanisms in Bahasa Indonesia. MITWPL 14, 183-218. Schachter, Paul. 1976. The Subject in Philippine Languages: topic, actor, actor-topic, or none of the above? In Charles Li (ed.) Subject and Topic. New York: Academic Press, 491-518. Schachter, Paul and Fe T. Otanes. 1972. Tagalog Reference Grammar. Berkeley: University of California Press. Sells, Peter. 2000. Raising and the order of clausal constituents in the Philippine languages. In I. Paul, V. Phillips and L. Travis (eds.) Formal Issues in Austronesian Linguistics. Dordrecht, Kluwer Academic Press, 117-144. Proceedings of the 26th West Coast Conference on Formal Linguistics edited by Charles B. Chang and Hannah J. Haynie Cascadilla Proceedings Project Somerville, MA 2008 Copyright information Proceedings of the 26th West Coast Conference on Formal Linguistics © 2008 Cascadilla Proceedings Project, Somerville, MA. All rights reserved ISBN 1-57473-423-2 library binding A copyright notice for each paper is located at the bottom of the first page of the paper. Reprints for course packs can be authorized by Cascadilla Proceedings Project. Ordering information Orders for the library binding edition are handled by Cascadilla Press. To place an order, go to www.lingref.com or contact: Cascadilla Press, P.O. Box 440355, Somerville, MA 02144, USA phone: 1-617-776-2370, fax: 1-617-776-2271, e-mail: sales@cascadilla.com Web access and citation information This entire proceedings can also be viewed on the web at www.lingref.com. Each paper has a unique document # which can be added to citations to facilitate access. The document # should not replace the full citation. This paper can be cited as: Gerassimova, Veronica and Peter Sells. 2008. Long-Distance Dependencies in Tagalog: The Case for Raising. In Proceedings of the 26th West Coast Conference on Formal Linguistics, ed. Charles B. Chang and Hannah J. Haynie, 190-198. Somerville, MA: Cascadilla Proceedings Project. or: Gerassimova, Veronica and Peter Sells. 2008. Long-Distance Dependencies in Tagalog: The Case for Raising. In Proceedings of the 26th West Coast Conference on Formal Linguistics, ed. Charles B. Chang and Hannah J. Haynie, 190-198. Somerville, MA: Cascadilla Proceedings Project. www.lingref.com, document #1672.