Immobile wh-phrases in Tagalog

I argue that overt movement cannot always be triggered by a need to create a specifier for the head bearing the Probe, as commonly assumed. In Tagalog, I claim, the best description of the behavior of wh-phrases is that they must become linearly adjacent to C; I show that Tagalog wh-phrases in fact occupy several different structural positions which are consistent with this linear requirement.


Introduction
This paper will bring some new data to bear on the question of how "overt movement" is to be characterized.I will argue that Probes which trigger overt movement cannot require movement to a position c-commanding them, as standardly assumed.The account will be consistent with accounts like that of Richards (2010;2016), which only require that movement land in a position which is linearly close to the triggering Probe.
The data will all come from Tagalog.We will see, first, that Tagalog wh-movement is generally obligatorily overt.However, there is a class of wh-phrases in Tagalog (wh-predicates), which I will argue are incapable of overt wh-movement.We will see that wh-predicates can form wh-questions just if they are linearly adjacent to the interrogative complementizer which begins the clause, and that they are never in the structural position occupied by other kinds of wh-moved phrases.

Wh-predicates in Tagalog
Tagalog is a predicate-initial language.Predicates of any lexical category are typically initial in the clause: (1) a. Na-sa unibersidad ang lola ninyo pred-dat university ang grandmother your.pl'Your grandmother is at the university' b.Mabilis ang butiki fast ang house.lizard'The house lizard is fast' c.Pangulo si Duterte president ang Duterte 'Duterte is president' d.Bibilh-in ni Juan ang mga libro-ng ito fut.buy-acc unm Juan ang pl book-li this 'Juan will buy these books' And wh-questions, whether matrix or embedded, may have initial predicates as their wh-phrases: (2) a. Na-saan ang lola ninyo?pred-where ang grandmother your.pl'Where is your grandmother?' b.Gaano kabilis ang butiki?how fast ang house.lizard'How fast is the house lizard?'The boldfaced material in (2-3) actually has two reasons to be initial in the Tagalog clause.Not only are these predicates, in a generally predicate-initial language, but they are wh-phrases, and wh-movement is obligatory in Tagalog: 3 (4) a. *Pumunta saan si Maria?nom.wentwhere ang Maria 'Where did Maria go?' the predicate alam 'know', rather than after it as a non-pronominal DP would. 2 I have glossed the complementizer kung with C mvmt , which is intended to indicate that it is a complementizer that triggers movement of an operator to itself.This is not the complementizer used in ordinary declarative clauses, but it is used in embedded questions, in relative clauses with non-DP relative operators (see Otsuka and Tanaka 2016;Hsieh 2018 for further discussion), and in adjunct clauses with meanings like if or when (and see Larson 1985;Enç 1987;Bhatt and Pancheva 2006, along with much other work, for arguments that if and when-clauses in other languages involve operator movement). 3Here there is apparently some cross-speaker variation: Law and Gärtner (2005) describe speakers of Tagalog who accept (4b).Speakers that I have consulted have rejected it.All that will be important for the claims to be made here is that speakers who reject (4b) exist; speakers with the grammar described by Law and Gärtner will also be consistent with the paper's claim, though they will not offer the same kind of evidence for it.

b. *Bumili
ng isda ang alin-g babae?nom.bought unm fish ang which-li woman 'Which woman bought fish?'But in fact, we will see evidence that wh-predicates in Tagalog are incapable of wh-movement; they must remain in situ.One piece of evidence for this conclusion comes from long-distance wh-questions.It turns out that wh-predicates in Tagalog cannot be extracted long-distance: ( Tagalog wh-predicates, then, are subject to a special restriction; unlike other wh-phrases in Tagalog, they cannot move across clause boundaries. 5Demonstrating that wh-predicates cannot wh-move in monoclausal questions in Tagalog is harder to do, since predicates and wh-phrases are both typically clause-initial.I turn to that task in the next section. 4It is unlikely that the moving phrase in (6b) is actually the DP ano 'what'; DP wh-questions in Tagalog generally involve some kind of cleft construction, so the moving phrase would be a (null) relative operator, yielding a structure something like that of 'What is the (thing which) the teacher said that Juan would read?'.See Richards (1998), Aldridge (2002), Mercado (2004), Hsieh (2020) for further discussion. 5A reviewer asks how the meanings in (5) are expressed.Speakers pressed on this issue typically offer paraphrases with structures and meanings like that of "What will Juan do with the books, according to the teacher?".

Wh-predicates in monoclausal wh-questions
Much of the literature on Tagalog agrees that predicate-initial word order is created by movement of the predicate to a syntactic position higher than the arguments of the clause (Kroeger 1993;Richards 1993;2000;Rackowski 2002;Aldridge 2004;and much other work).Certain kinds of material can intervene between the fronted predicate and overt complementizers, indicating that the landing site of predicate fronting is not C: Sinabi ni Maria [ na hindi naintindih-an ng propesor ang tanong niya] acc.said unm Maria C not understood-dat unm professor ang question her 'Maria said that the professor did not understand her question' In ( 6), the predicate of the embedded clause, naintindihan 'understood', is separated from the embedding complementizer na by the negative morpheme hindi.Data like ( 6) have been taken to argue (for example, by Richards 2003 andAldridge 2004), not only that the landing site of predicate fronting is not C, but that it should not be the sister of C, either; there must be a certain amount of syntactic space between C and the position of the predicate.
Negation is not the only kind of material that can occupy this space.Certain kinds of phrases may be fronted to positions before the predicate, where they are followed by an invariant morpheme ay.Thus, for example, the ordinary word order in (8a) alternates with the inverted one in (8b), and the resulting inverted structure can also appear in embedded clauses, as in In embedded wh-questions, the order in ( 11) is the only possible one for these elements.In particular, an ay-fronted phrase cannot intervene between the interrogative complementizer kung and the fronted wh-phrase: We can think of (12) as an illustration of what it means to say that Tagalog has obligatory wh-fronting; wh-fronting is to a position just to the right of the embedding complementizer kung, and the wh-phrase cannot be separated from this position.We can update the description in (10) as follows: The description in (13) is not exhaustive; there are other kinds of things that can precede the Tagalog predicate, such as fronted topics and certain kinds of adverbs.But this description will suffice for our purposes.
The fact that the wh-phrase follows the complementizer kung rather than preceding it is surprising on conventional assumptions about the architecture of the clause; if wh-movement lands in the specifier of CP, we should expect the wh-phrase to precede the complementizer.
The literature contains several plausible proposals about this problem, any of which would be consistent with the point being made here.Kroeger (1993), Otsuka and Tanaka (2016), and Hsieh (2020) claim that Tagalog wh-movement is to the specifier of a projection below the one headed by kung.Sabbagh (2014) posits wh-movement to the specifier of kung, together with conditions on how Tagalog syntactic structure is mapped onto prosodic structure which guarantee that the wh-phrase must be linearized after the complementizer.
More generally, a natural question to ask at this point would be how the facts in (13) should be represented in a tree.We might consider one extreme in which each of the elements involved occupies its own dedicated maximal projection: In a structure like the one in (14b), the facts in (16-18) are easy to describe: Tagalog can elide ayP and TP, but not NegP. 7In (15b), on the other hand, description is less straightforward. 7As a reviewer points out, we must also exclude a representation of (18) involving ellipsis of the TP complement of a null (positive) version of Neg.As it happens, ellipsis of the null complement of Neg must independently be ruled out; a version of (17) in which hindi 'not' is removed is ungrammatical (with the intended meaning '…but Mary said that I did').
Whether we want to posit all the structure in (14b) is unclear, but at least with our current understanding of how ellipsis is licensed, it is hard to see how the structurally impoverished structure in (15b) can be the right one.
We now seem to have found a kind of case in which the position of the predicate and the position of a fronted wh-phrase can be linearly (and, I have just argued, structurally) distinguished; material can intervene between these two positions.Where does a wh-predicate appear?
The fact is that wh-predicates are incompatible with ay-inversion in embedded questions: Neither of the orders in ( 19) is possible. 8We can relate the ill-formedness of (19a) to the contrast in (11)(12), repeated here as ( 20 As (20) reminds us, we have already seen that wh-phrases cannot be separated from the complementizer by ay-inverted phrases.But in (19b), we can see that fronting the wh-predicate across the ay-inverted phrase is also unacceptable.The ill-formedness of (19b) is reminiscent of the facts in (5), repeated as ( 21 unm Juan ang pl book-li this 'What did the teacher say Juan would do with the books?' Wh-predicates, like other wh-phrases, must be adjacent to the interrogative complementizer, but in every case in which fronting a wh-predicate would not be string-vacuous, fronting is blocked; they cannot front across clause boundaries, as (21) shows, but they also cannot front within their clause, as we see in (19). 9 The facts in matrix wh-questions are somewhat more complicated.Here it is possible for ay-inverted phrases to precede wh-predicates: 10 (22) Ang Diyos mo ba ay na-saan?ang God your q ay pred-where 'Where is your God?' (Psalms 42:10, Magandang Balita Bible) We might understand the fact in ( 22) as telling us that Tagalog is like many languages in having kinds of movement to the left periphery which are restricted to root clauses; in this particular case, we would say, ay-inversion takes place to a higher position than it could in an embedded question, landing above the (null) complementizer and allowing the linear adjacency requirement between C and the wh-predicate to be satisfied. 119 I have only shown this for fronting across ay-fronted constituents, and not for negation.Wh-fronting of predicates across negation is independently ruled out, perhaps as an instance of the inner island effect of Ross (1984), properly understood: (i) *Where isn't your grandmother?
(ii) *Which president do you not think [Mary will be]? 10Matrix non-predicate wh-phrases can also be preceded by ay-inverted material: (i) Ngunit ang karunungan ay saan ba nagbubuhat?but ang wisdom ay where q nom.come.from'But where does wisdom come from?' (Job 32:8, Magandang Balita Bible) 11 I have omitted the question particle ba; like the pronoun mentioned in footnote 2 above, ba is a clitic, so its placement in the sentence is not entirely determined by the syntax.The second movement step in ( 23) would be the one which is banned in embedded clauses but licit in matrix questions. 12The question of why such root phenomena should exist, in Tagalog or any other language, is a very interesting one that lies beyond the scope of this paper; 13 for further discussion of root phenomena in other languages, see Emonds (1969;2004), Bianchi and Frascarelli (2010), Haegeman (2012) 12 Alternatively, we might handle this kind of root phenomenon by allowing ayP to be generated above CP just in root clauses. 13In an approach like that of Richards (2010;2016), in which the placement of the wh-phrase is determined by constraints imposed by the interface with PF, it is tempting to build an explanation around the fact that in Tagalog (as in many other languages) matrix C is reliably phonologically null, unlike embedded C. Since wh-movement in matrix questions is obligatory in Tagalog, however, we cannot simply say that this null C is free of any conditions on its relation with the wh-phrase, at least if we are to rely on these conditions to drive wh-movement.
We have now seen evidence that the boldfaced wh-phrases in (24) are in different syntactic positions.Saan 'where' in (24a) is in a comparatively high structural position, preceding the positions of negation and ay-inverted phrases.The predicate wh-phrase na-saan 'pred-where' in (24b), on the other hand, is in a comparatively low structural position, the predicate-initial position typically occupied by predicates.In fact, we have seen that wh-predicates cannot move to any position higher than the one generally occupied by predicates. 14  The description in (25) covers the data under discussion: (25) In Tagalog, wh-phrases must be linearly adjacent to C.
On this view, the wh-phrase in (24a) satisfies ( 25) via actual wh-movement, while the wh-phrase in (24b) satisfies ( 25) because of the general processes that typically make predicates initial in Tagalog.In fact, we have now seen that if a wh-predicate cannot satisfy (25) while staying in the ordinary position of predicates, then ( 25) is violated and the result is ill-formed; wh-predicates cannot undergo the kind of wh-movement that we see in (24a).
The condition in (25) would be a natural consequence of an approach to overt movement like that in Richards (2010;2016).In that theory, overt movement takes place because of conditions that require Probes and Goals to share prosodic structures of certain kinds.I claim that the construction of prosodic structure begins earlier in the derivation than we have classically believed, and that operations in the narrow syntax can be driven by the need to improve the prosodic representation.The approach posits a requirement that the smallest string containing C and the wh-phrase have the wh-phrase as its most prosodically prominent element.In languages like Tagalog, I claim, the most prosodically prominent element is the leftmost phrase in a given string. 15The upshot of this would be that in the smallest string containing C and the wh-phrase, the wh-phrase must be the leftmost phrase.A wh-phrase which immediately follows C will obey this condition, regardless of how much syntactic structure intervenes between them, as long as no phrase is pronounced between C and the wh-phrase.On this kind of account, then, we expect to find generalizations like (25).
14 I will refer interested readers to Richards (2009) for some discussion of why this might be.Briefly, I propose there that Tagalog nonverbal predicates involve a null copula, which cannot be stranded by movement (I compare the Tagalog null copula to the AAVE one in (ib) (Labov 1969)): (i) a.He Ø fast in everything he do.
b. *How beautiful you Ø ! 15 This is, for me, one setting of a binary parameter; there are also languages in which the most prosodically prominent element is the rightmost phrase.fronted wh-phrases and wh-predicates do not occupy the same syntactic position (the ay-inverted phrase intervenes between them), but both can (and must) satisfy the condition on wh-phrases in Tagalog.An account of the standard kind, which says that wh-fronting is obligatory in some languages because a particular head demands that the wh-phrase land in its specifier, appears to fail for Tagalog.Overt wh-movement is obligatory in Tagalog, but there is no single structural position occupied by all Tagalog wh-phrases.We are used to saying that wh-movement is driven by a need to fill the specifier of CP, but I have argued that this is a mistake.Wh-movement, if I am right, is driven by a need to make sure that a particular domain has a wh-phrase at its linear left edge, regardless of where that left edge appears in syntactic structure.
(12) *Alamin muna kung ikaw ay kailan hihinto.acc.know first C mvmt ang.you ay when nom.fut.stop'Figure out first when you will stop' first C mvmt ang.you ay pred-where now 'Figure out first where you are now' b. *Alamin muna kung na-saan ikaw ay ngayon know.accfirst C mvmt pred-where ang.you ay now 'Figure out first where you are now' tumaya kung kailan ikaw know.accfirst before ang.you nom.start-li nom.betC mvmt when ang.you ay hihinto.ay nom.fut.stop'Figure out first before you start betting when you will stop' b. *Alamin muna kung ikaw ay kailan hihinto.know.accfirst C mvmt ang.you ay when nom.fut.stop'Figure out first when you will stop'

what ang said unm teacher C fut.read-acc unm Juan 'What did the teacher say that Juan would read?'
Schachter and Otanes (1972)lipina 'I don't know whether Maria is Filipina' I will refer to the process in (8b-c), followingSchachter and Otanes (1972), as ay-inversion.It is generally optional, and if it has effects on information structure, they are quite subtle.first before ang.you nom.start-li nom.betC mvmt when ang.you ay hihinto.nom.fut.stop'Figure out first before you start betting when you will stop' (http://www.casinopatnubay.com/pagseset-ng-budget/,accessed 5/22/20) know C mvmt why ang Maria ay not nom.left'Idon't know why Maria didn't leave' b.A logically possible opposite extreme would posit a single projection hosting all of the material after C, perhaps with the predicate occupying the head of the projection, and the rest of the material in various adjunct and specifier positions: Maria ay not nom.leftbut ang Juan ay nom.left 'Maria didn't leave, but Juan did leave'