1. Introduction
In this article, I develop a new analysis of partial wh-movement in Indonesian which resolves apparent challenges posed by this construction for criterial freezing (hereafter, CF) (Rizzi Reference Rizzi, Cheng and Corver2006, Reference Rizzi and Phoevos Panagiotidis2010, Reference Rizzi and Shlonsky2015, Reference Rizzi2016, Reference Rizzi2017).Footnote 1 I propose that this construction involves a multi-criterial configuration, wherein an XP endowed with a criterial feature F1 (focus) properly contains a YP endowed with another criterial feature F2 (Q); the partial movement structure is derived when the XP undergoes focus movement into the specifier of an embedded non-interrogative CP, followed by sub-extraction of the YP contained within the XP to move to the specifier of the matrix interrogative CP. The proposed analysis leads to the expectation that there should be other cases of multi-criterial configurations involving focus-/wh-movement created by sub-extraction. I will show that this expectation is borne out by Japanese wh-questions additionally marked with the focus particle sae ‘at least’ (Maeda Reference Maeda2019).
The paper is organized as follows. In section 2, I review Cole and Hermon's (Reference Cole, Hermon, Lutz, Müller and von Stechow2000) pseudo-cleft analysis of partial wh-movement in Indonesian and highlight that such a derivation entails a multi-criterial movement − overt focus movement, followed by covert wh-movement − a configuration that should be erroneously ruled out by CF. In section 3, I propose that the derivation of this construction employs sub-extraction, or movement out of a moved element, to evade a freezing violation by means of a two-layered configuration involving partially moving wh-phrases in this language. According to this analysis, such wh-phrases are made up of two layers, with the focus layer immediately dominating the Q layer. The partial movement structure is obtained when the embedding focus phrase undergoes focus movement to an intermediate C-system, thereby “sheltering” the embedded QP from incurring a freezing violation, followed by sub-extraction of the QP from within the focus phrase to target the matrix interrogative C-system. In section 4, I show that my new approach to multi-criterial movement is independently supported by cases of Japanese wh-questions endowed with an additional focus feature, as documented in Maeda (Reference Maeda2019). I conclude in section 5 with a brief discussion of some remaining issues.
Unless otherwise indicated, all the Indonesian data in this paper were collected by the author from two Indonesian speakers who are both familiar with acceptability judgement tasks used in syntax. One speaker is Javanese from Kendal, Central Java and has used Indonesian for communication and schools since childhood, together with Central Javanese (her mother language). The other speaker uses a variety of Peranakan Javanese spoken in Malang to communicate with his parents and relatives. He has learned Standard Indonesian from kindergarten to high school and is familiar with Colloquial Jakarta Indonesian from TVs and newspapers as well as from his friends; he lived in Jakarta as an undergraduate student for around eight months in 2004–2005. The author created the relevant Indonesian examples on his own first and later had the speakers check their acceptability based on their judgements of what would constitute some “standard” version of Indonesian.
2. The partial wh-movement paradox in Indonesian
Indonesian has three options for forming wh-questions (Saddy Reference Saddy, Cheng and Demirdache1991; Cole and Hermon Reference Cole and Hermon1998, Reference Cole, Hermon, Lutz, Müller and von Stechow2000; Cole et al. Reference Cole, Hermon, Aman, Gil and Collinsto appear): full wh-movement, partial wh-movement, and wh-in-situ. These options are illustrated in (1a−c).
Cole and Hermon (Reference Cole, Hermon, Lutz, Müller and von Stechow2000) propose that partial movement of nominal wh-phrases with yang in Indonesian/Malay consists of two movement steps: focus movement of a nominal wh-phrase to an intermediate CP, followed by further movement of the same phrase to the matrix CP. To illustrate their analysis, let us start by considering an example of a nominal wh-question as in (2).
Cole and Hermon argue that (2) has a pseudo-cleft sentence shown in (3) as its derivational base, which differs minimally from a regular non-wh cleft construction in (4a), in that the focused constituent is the wh-phrase apa ‘what’ in the former, but the focused phrase buku ‘book’ in the latter. Examples (4a—b) show that fronting of the focus phrase to the sentence-initial position is optional.
Based on these parallels between nominal wh-questions with yang and regular cleft constructions, Cole and Hermon argue for a pseudo-cleft derivation for the former. According to this analysis, the relevant part of the derivation for the nominal wh-question in (2) is as shown in (5).
(5)
The derivation in (5) involves two NPs. One is the headless relative clause subject introduced by the relative complementizer yang. Within this NP, the null operator undergoes movement from its base thematic position to the specifier of the local CP. The other NP is a nominal expression, an in-situ wh-phrase. The headless relative clause subject and the wh-phrase, in turn, are connected by the null variant of the optional copula adalah ‘be’ (see (3) and (4a)). The surface wh-initial word order is derived by overt focus movement of the wh-phrase to the specifier of the matrix CP.
Given Cole and Hermon's (Reference Cole, Hermon, Lutz, Müller and von Stechow2000) analysis, then, the derivation of partial (nominal) wh-movement at least involves focus movement to an intermediate C-system, as depicted in (6).
(6)
The focus movement analysis of partial wh-movement in Indonesian is indirectly supported by Kikuyu, in which a wh-phrase can undergo partial movement only if it is prefixed with an overt focus particle. For example, in (7), the wh-phrase o ‘who’ combines with the focus particle ne to form nóo. See also Muriungi (Reference Muriungi2005) and Abels (Reference Abels2012) for the same observation in Kȋȋtharaka.
Rizzi (Reference Rizzi, Cheng and Corver2006, Reference Rizzi and Phoevos Panagiotidis2010, Reference Rizzi and Shlonsky2015, Reference Rizzi2016, Reference Rizzi2017) holds that a focus head counts as a criterial position and argues that an XP moved to a position associated with some scope/discourse properties (e.g., Q, Foc, Top, relative) must stop in that position.Footnote 2 This principle, known as CF, is defined as shown in (8).
(8) Criterial Freezing (first version): A phrase meeting a criterion is frozen in place. (Rizzi Reference Rizzi, Cheng and Corver2006: 112)
The derivation in (6) means, then, that the wh-phrase undergoes criterial movement to the specifier of the intermediate focus head and must stop there as the result of CF.
However, there is convincing evidence, first documented by Saddy (Reference Saddy, Cheng and Demirdache1991) (see also Cole and Hermon Reference Cole and Hermon1998, Reference Cole, Hermon, Lutz, Müller and von Stechow2000 for further data), that a wh-phrase supposedly frozen in an embedded C-system, as in (6), undergoes further criterial wh-movement into the matrix C-system. Saddy's evidence supporting this additional covert wh-movement comes from island effects shown in (9) and (10).
Example (9a) is ungrammatical because the overt wh-movement of siapa ‘who’ crosses the subject island. More telling is the ungrammaticality of (9b) despite the fact that, unlike in (9a), the wh-phrase does not seem to cross the island boundary, at least overtly. As a point of comparison, the grammaticality of (9c) shows that an in-situ wh-phrase is insensitive to the subject island effect. In the case of the absence of island effects in wh-in-situ questions in Indonesian/Malay, Cole and Hermon argue that “the (wh-OP) question operator is merged at the root Spec CP, and, therefore, unselectively binds a wh-variable in its scope” (Reference Cole and Hermon1998: 240), following Tsai (Reference Tsai1994) and Reinhart (Reference Reinhart2006). Saddy (Reference Saddy, Cheng and Demirdache1991) argues that the ungrammaticality of (9b) is accounted for if the partially moved wh-phrase undergoes covert wh-movement into the matrix CP, triggering the subject island violation. A similar argument for covert wh-movement can be made on the basis of the data in (10a−c) with regard to adjunct island effects.
In fact, a scope interaction between a quantifier in the matrix clause and the partially moved wh-phrase in Indonesian presents further support for the covert wh-movement step. In German, another partial wh-movement language, a partially moved wh-phrase cannot take scope over elements in the matrix clause in LF. This observation is illustrated in (11), for which Pafel notes that “wide scope of the universal quantifier seems to be the only option” (Reference Pafel, Lutz, Müller and Stechow2000: 340). The impossibility of the wide scope reading of the wh-phrase wo ‘where’ with respect to the universal quantifier jeder ‘everyone’ in the matrix clause is indeed confirmed by the semantic anomaly of the example in (12), which is so constructed to fit a small discourse which forces this particular scope construal (wh>∀).Footnote 3
Partial wh-movement in Indonesian behaves differently, for Saddy (Reference Saddy, Cheng and Demirdache1991) points out that in (13), the partially moved wh-phrase may take scope over the universal quantifier in the matrix clause.
The scope pattern illustrated in (13) then lends further support to the covert wh-movement analysis for partial wh-questions in Indonesian.
In light of the two observations from Saddy (Reference Saddy, Cheng and Demirdache1991) on island-sensitivity and scope facts, the full derivation of partial wh-movement in Indonesian must now be as shown in (14) instead of (6).
(14)
However, (14) involves a multi-criterial configuration where a single wh-phrase enters into more than one criterial relationship with two discourse-functional heads (i.e., Foc and Q) in the left periphery of the two successive clauses, seemingly violating the CF. In the next section, I will develop a proposal to resolve this challenge raised by partial wh-movement in Indonesian for CF.
3. The partial wh-movement paradox, sub-extraction, and multi-criterial movements
My analysis builds on what one might call the two-layered hypothesis for wh-/focus-phrases developed by Cable (Reference Cable2007) and Horvath (Reference Horvath, Bennis, Everaert and Reuland2000, Reference Horvath, Karimi, Samian and Wilkins2005). I will briefly outline this hypothesis in section 3.1 to lay the groundwork for my analysis of Indonesian partial wh-movement in section 3.2.
3.1. The two-layered hypothesis for wh-/focus-phrases
Cable (Reference Cable2007) proposes that the interrogative C head probes for an interpretable Q-feature of the Q-particle, not any feature of the wh-word itself, contrary to the standard generative assumption. In this theory, schematically depicted in (15), the C head agrees with the QP, and this agreement, in turn, triggers movement of the QP into the specifier of the CP.
(15)
In a similar vein, Horvath (Reference Horvath, Bennis, Everaert and Reuland2000, Reference Horvath, Karimi, Samian and Wilkins2005) argues that so-called ‘focus movement’ in Hungarian is triggered by the Exhaustive-Identification Operator (EI-OP), a phonologically null variant of the overt association-with-focus particle cask ‘only’, and hence has nothing to do with the focus feature itself. According to this analysis, ‘focus movement’ in this language is analyzed as overt movement of an XP containing the EI-OP operator, not of the FocP itself, as schematically represented in (16).
(16)
This analysis is supported by the observation that focus movement to the immediately preverbal position in Hungarian is available only if the target of the movement is construed as exhaustively identifying the true answer to a question. To illustrate this observation, consider (17−18).
In (17), the dative phrase is understood to present the exhaustive answer to the wh-question, and is subject to overt focus movement to the immediately preverbal position, as shown by the contrast in grammaticality between (17a) and (17b). The postpositional phrase in (18), by contrast, is not similarly understood and thus stays in its base thematic position instead of undergoing focus movement.
3.2. Criterial freezing, sub-extraction, and multi-criterial movement in Indonesian
Let us hypothesize that what appears to be a wh-phrase in Indonesian partial wh-movement is actually made up of two layers: the focus layer and the Q layer (see also Sabel (Reference Sabel, Lutz, Müller and Stechow2000) for a similar proposal that wh-movement involves checking both [+wh] and [+focus] features, and section 5 for a relevant discussion). I propose that partial wh-movement in Indonesian is obtained when the FocP undergoes overt focus movement into an embedded C-system, followed by covert sub-extraction of the QP contained within the FocP to the matrix C-system, as depicted in (19).
(19)
The two-step derivation in (19) successfully overcomes the problem identified in section 2. The FocP undergoes a criterial movement into [Spec, CP2] and is frozen in place as the result of CF. The FocP contains a QP endowed with Q-feature; the QP, then, is sub-extracted from within the FocP and undergoes another criterial movement into [Spec, CP1] and stops there.
The proposed analysis has close affinities with Abels's (Reference Abels2012) theory of partial wh-movement. According to his theory, wh-phrases in partial wh-movement languages such as Kȋȋtharaka and Indonesian/Malay are externally merged with a null operator, O[Foc][u Wh↓↑], which has a valued focus feature and an unvalued wh-probe; the operator in question first undergoes focus movement, pied-piping the wh-phrase to an intermediate CP, and subsequently undergoes wh-movement on its own, obligatorily stranding the wh-phrase, as schematically depicted in (20).Footnote 4
(20)
In Abels's framework, the notation [uF↓↑] denotes a feature whose sharing requires mutual c-command between a syntactic head H and a different syntactic object O (Abels Reference Abels2012). In (20), C2[Foc↓↑] and C1[wh↓↑] are designed to implement focus movement and wh-movement, respectively. It is to be noted, however, that the derivation in (20) violates CF in that the same operator first undergoes focus movement to the embedded clause and then undergoes wh-movement to the matrix clause. This is precisely the situation that is circumvented by the sub-extraction analysis in (19).
I wish to address one issue at this point. As is clear from (19), the relevant derivation violates the traditional ban on movement out of moved elements (Ross Reference Ross1967, Reference Ross, Weimer and Palermo1974; Postal Reference Postal1972; Culicover and Wexler Reference Culicover, Wexler, Culicover, Wasow and Akmajian1977; Collins Reference Collins1994; Takahashi Reference Takahashi1994; Müller Reference Müller1998, Reference Müller2010; Boeckx Reference Boeckx2008; Gallego Reference Gallego2009; Uriagereka Reference Uriagereka2012), as subsequent wh-movement of the QP is launched from within the moved FocP.Footnote 5 However, Bošković (Reference Bošković2018, Reference Bošković, Belletti and Collins2021) argues that the traditional ban holds only for successive-cyclic movement out of a moved element, and presents evidence that elements that are either base-generated at, or moved to, the edge of a moved element, independently of successive-cyclic movement, can actually undergo sub-extraction − a generalization that he derives from the interaction of the Phase Impenetrability Condition (PIC) and labelling (Chomsky Reference Chomsky2013, Reference Chomsky, Domenico, Hamann and Matteini2015). Consider (21) and (22). Serbo-Croatian example (21) illustrates that the possessor XP, base-generated at the edge of the YP, can move out of the YP (the TNP stands for the traditional NP). Example (22), from Dutch, shows that the R-pronoun, standardly assumed to move to the specifier of the PP, may undergo movement out of the relevant phrase even if the latter itself is moved. Both cases are grammatical despite the fact that movement of the possessor phrase/R-pronoun violates the freezing ban, unlike in (23), where the derivation involves successive-cyclic movement of the wh-phrase out of the DP to the specifier of the interrogative CP.
Bošković (Reference Bošković2018, Reference Bošković, Belletti and Collins2021) argues that the generalization above can be deduced as follows. In any situation where the wh-phrase must move to the edge of the DP phase in a successive-cyclic fashion due to the PIC, as exhibited in (23) the relevant movement does not result in labelling of the syntactic object so formed due to the lack of feature-sharing. This then de-phases the syntactic object on top of the DP phase, deriving the freezing effect observed in (23) under the assumption that only phases can undergo movement (Chomsky Reference Chomsky, Martin, Michaels and Uriagereka2000, Reference Chomsky and Kenstowicz2001; see also Matushansky Reference Matushansky, McGinnis and Richards2005, Rackowski and Richards Reference Rackowski and Richards2005, J. H.-T. Cheng Reference Cheng2012, Harwood Reference Harwood2013, Legate Reference Legate2014 and Bošković Reference Bošković2015). On the other hand, if an XP is base-generated at, or moved to, the edge of YP, independently of the PIC, as illustrated in (21) and (22), that means that merger of the XP and the YP can involve feature-sharing for labelling, so that the new syntactic object created by this merger is not delabelled. It follows that movement of the XP can successfully be launched from the already moved YP. The proposed deduction, then, replaces the traditional freezing ban with the new generalization in (24), which, in turn, can be reformulated as in (25) within the labelling framework.
Importantly, Bošković (Reference Bošković, Belletti and Collins2021) points out that the smuggling derivation of the kind proposed by Collins (Reference Collins2005a, Reference Collins2005b) along the lines shown in (26) is in compliance with his revised phase-theoretic/labelling-based replacement of the traditional freezing ban if the direct object this book does not undergo movement into [Spec, PartP], as argued by Collins himself, but instead moves out of the PartP directly to [Spec, TP], as schematically depicted below.Footnote 6
(26)
This derivation suggested by Bošković (Reference Bošković, Belletti and Collins2021) is not blocked by (24) or (25), for the PartP is not only a non-phase head (i.e., VoiceP is the phase head for Collins Reference Collins2005a), but also lacks a specifier in the first place, thereby making the two conditions irrelevant.
Notice, however, that Bošković's deduction of the revised generalization in terms of the interaction of the PIC with labelling actually disallows the smuggling derivation shown in (26) because it is based on the assumption that only phases can undergo movement; the PartP, a non-phasal object, undergoes movement in this derivation. For this reason, Bošković (Reference Bošković2018, Reference Bošković, Belletti and Collins2021) actually puts forth a new deduction of the conditions in (24) and (25) which nonetheless does not block movement of non-phases. Let us assume that movement is triggered by an uninterpretable/unvalued feature of the moving element (Bošković Reference Bošković2007, Reference Bošković, Lima, Mullin and Smith2011) and that the movement of a syntactic object resulting from the merger of X and Y requires it to be labelled so that either X or Y may project and pass the relevant feature to the object. These assumptions have the desired consequence that unlabelled elements cannot undergo movement, allowing the generalization, but without the proviso that only phases are entitled to movement. With these assumptions in place, it is easy to see that my proposed derivation for partial wh-movement in Indonesian in (19) is entirely parallel to Bošković's rendition of the smuggling derivation for the passive construction in (26): The FocP moves to the specifier of the embedded C-system/the Foc head. The QP contained within the FocP then undergoes direct movement out of it without passing through its specifier to the specifier of the matrix interrogative C-system. Consequently, this derivation is in accordance within Bošković's (Reference Bošković2018, Reference Bošković, Belletti and Collins2021) replacement of the traditional ban on movement out of moved elements.
3.3. A challenge from partial wh-questions in Indonesian without the focus marker yang
As noted by Cole and Hermon (Reference Cole, Hermon, Lutz, Müller and von Stechow2000) and Cole et al. (to appear), non-nominal wh-phrases such as kenapa ‘why’ and bagaimana ‘how’ cannot co-occur with yang, as shown in (27), but nonetheless may still undergo partial wh-movement, as illustrated by the grammaticality of (28).
Example (27) appears to indicate that the derivation of (28) cannot be based on the pseudo-cleft construction with optional focus movement. Indeed, the headless relative clause option, available for nominal wh-phrases, is blocked for non-nominal wh-phrases, as shown in (29). Note, furthermore, that island effects are observed between the overt landing site of the adjunct wh-phrase and its intended scope position, as shown in (30), suggesting that covert wh-movement indeed takes place.
Cole and Hermon conclude that the derivation of non-nominal partial wh-questions like (29) should be as shown in (31) and suggest that the overt movement to [Spec, CP2] is triggered not by the focus feature, but instead by the need of the moved wh-operator itself, namely, that “a question operator must be located in the specifier relationship with a complementizer” (Reference Pafel, Lutz, Müller and Stechow2000: 109).Footnote 7
(31)
Note that just because the pseudo-cleft derivation is unavailable for non-nominal partial wh-questions does not mean that the overt movement step of their derivation cannot be focus-driven. In fact, I will argue below that my analysis can be maintained for such wh-questions as well. Suggestive evidence for this position comes from Fortin's (Reference Fortin, Paul, Chung, Finer and Potsdam2009) observation that Indonesian is a “unique focus language” in the sense of Stoyanova (Reference Stoyanova2008). Rizzi (Reference Rizzi and Haegeman1997) observes that Italian blocks multiple focus phrases, as illustrated in (32a). He further notes that a wh-phrase can co-occur with a topic, but not with a focused phrase, as shown by the contrast in grammaticality between (32b) and (32c). The contrast falls into place if Italian allows only one focus position per clause and wh-movement targets [Spec, FocP].Footnote 8
Fortin (Reference Fortin, Paul, Chung, Finer and Potsdam2009) provides examples in (33a−c) to show that Indonesian behaves on a par with Italian. The language thus blocks multiple focus phrases within a single clause, as illustrated in (33a). It also blocks a focus phrase marked with the focus particle lah to co-occur with a wh-phrase, whether the latter undergoes movement or remains in-situ: see (34b, c). Note that a wh-phrase or a focused phrase can co-occur with a topic phrase in the same clause, as shown in (33d, e).
Now, given Fortin's observation, my analysis of yang-less partial wh-questions predicts that partial wh-movement should be blocked for non-nominal wh-phrases in the presence of another focused expression in the embedded clause. This prediction is borne out by the inability of a partially moved non-nominal wh-phrase as to co-occur with kemarin-lah ‘yesterday’, as shown in (34a).
Also expected is the fact, illustrated in (34b), that the full wh-movement variant is ungrammatical; the single focus condition is violated at the time when the adjunct wh-phrase stops in [Spec, CP2]. The new observation above, thus, provides support for the view that partial wh-movement of both nominal and non-nominal wh-phrases involves focus movement despite the fact that the derivation of the latter is not visibly based on the pseudo-cleft source headed by yang with focus movement.
If all Indonesian wh-phrases are contained within the FocP layer, one might wonder whether my proposed analysis would predict that multiple wh-questions should be unacceptable, given the single focus constraint noted above. However, Yanti (Reference Yanti2000) points out that Jambi, a local variety of Malay/Indonesian, allows multiple wh-questions, as shown in (35a−c).Footnote 9
Furthermore, I consulted two Indonesian speakers (see the final paragraph of section 1) regarding their own Indonesian counterparts to the Jambi Malay examples, given in (36a−c). They report that the examples in (36a) and (36b) are acceptable, though they both prefer the former to the latter. By contrast, (36c) is completely unacceptable to both of them.
The right empirical generalization to be drawn from (36a−c), then, seems to be that multiple wh-questions are grammatical in Indonesian as long as only one instance of overt wh-movement takes place to the interrogative CP, with the other wh-phrase(s) remaining in situ. Under my proposed system, this means that in-situ wh-phrases in the language are not associated with the focus feature, unlike partially moved wh-phrases, for otherwise such phrases should undergo focus movement. For this reason, I maintain, following Cole and Hermon (Reference Cole and Hermon1998, Reference Cole, Hermon, Lutz, Müller and von Stechow2000), that in-situ wh-phrases in Indonesian, lacking the multilayered focus/Q-structure, do not undergo any syntactic movement, but are rather variables licensed in situ by the base-generated wh-operator at the scopal [Spec, CP] through unselective binding/choice functions (Tsai Reference Tsai1994, Reinhart Reference Reinhart2006) (see section 2).Footnote 10
To sum up, I have argued that a uniform multilayered analysis can be maintained for partial movement involved in both nominal and non-nominal wh-questions regardless of whether such questions are accompanied by the focus marker yang. In fact, pursuing this analysis further, one might well propose that nominal wh-questions are in fact derived through the same mechanism as non-nominal wh-questions without necessarily assuming that they are derived from the underlying pseudo-cleft sentence. One indication that this analysis is on the right track comes from the observation that the overt realization of the copula adalah ‘be’ is impossible in the nominal wh-question with yang in (37) but optional in the pseudo-cleft sentence in (38) (repeated from (3)), which should be the source for (37) under Cole and Hermon's (Reference Cole, Hermon, Lutz, Müller and von Stechow2000) analysis of nominal wh-questions.
This discrepancy thus casts doubt on the idea of relating the two constructions in question in derivational terms through focus movement and indicates that (38) is indeed a pseudo-cleft with an in-situ wh-phrase, whereas (37) involves instead focus movement of the nominal wh-phrase without the pseudo-cleft source, just as in non-nominal wh-questions such as (27) and (28).Footnote 11
4. Further evidence for the multi-criterial focus/wh-movement from Japanese
I have argued in section 3 that sub-extraction from a double-layered complex wh-phrase in Indonesian creates a leeway for multi-criterial movement involving the satisfaction of two independent criterial features − Q and focus − for partial wh-movement in the language. In this section, I show that this analysis receives support from grammatical instances of wh-questions in Japanese endowed with an additional focus feature, recently documented in Maeda (Reference Maeda2019).
The Japanese particle -sae in (39a), which otherwise corresponds to even in English, may also yield the ‘at least’ interpretation in the antecedent of a conditional clause, as shown in (39b).
Kusumoto (Reference Kusumoto2001) argues that the sae-marked XP undergoes covert focus movement into the CP region headed by a conditional head on the basis of the contrast in grammaticality between (40) and (41). Example (40) is a baseline example to show that an XP marked by the particle may be associated with a conditional C head separated by more than one clausal boundary. The degraded status of (41) indicates that association of the sae-marked XP with the conditional C head is blocked by the adjunct island, which shows that the former undergoes covert movement into the specifier of the latter.
Turning now to the syntax of wh-questions in Japanese, examples in (42–44) are designed to show that this construction causes a CF effect through covert wh-movement. As pointed out by Miyagawa (Reference Miyagawa1998) and Yoshida (Reference Yoshida, Lin, Krause, Bruening and Arregi1999, Reference Yoshida2016), the two Q-particles in Japanese − nokai and ndai − can only be used for yes-no questions and wh-questions, as shown in (42) and (43), respectively.
Keeping this restriction in mind, consider (44). Here, the matrix clause is typed as a wh-question, as required by the Q-particle ndai, whereas the embedded clause is headed by the other Q-particle ka. The ungrammaticality of this example is accounted for if the wh-phrase undergoes covert movement into the embedded [Spec, CP] en route to the matrix [Spec, CP], triggering a CF violation.
We have seen thus far two types of criterial movement in Japanese: focus movement of an XP-sae on its ‘at least’ reading associated with a conditional C head, and covert Q-movement of an in-situ wh-phrase. Crucially for my present purposes, Maeda (Reference Maeda2019) observes that the wh-expression additionally marked with the focus particle, i.e., nani-sae, can receive two criterial interpretations in (45) − the Q-interpretation and the focus-interpretation − without any loss of grammaticality.
This example indicates that Japanese instantiates multi-criterial movement involving wh-/focus-features. I take Maeda's finding as additional support for the creation of a multi-criterial configuration through sub-extraction, just as what I have proposed for partial wh-movement in Indonesian. More specifically, the relevant part of the derivation for (45) should now be as depicted in (46). The wh-phrase nani-sae is made up of the focus layer dominating the Q-layer, just like a partially moved wh-phrase in Indonesian. The FocP undergoes focus movement into the CP2 and is frozen in place. However, the FocP has shielded its contained QP through this movement process so that the latter may evade a CF violation. Subsequently, the QP layer undergoes sub-extraction and covert movement into CP1 to check the Q-feature against its local interrogative C head.Footnote 12
(46)
5. Conclusions and Open Issues
In this paper, I have developed a new analysis of partial wh-movement in Indonesian that resolves seemingly challenging problems for CF. The analysis adopts the spirit of Cable's (Reference Cable2007) and Horvath's (Reference Horvath, Bennis, Everaert and Reuland2000, Reference Horvath, Karimi, Samian and Wilkins2005) multi-layered approach to wh-/focus-questions and proposes that the construction involves focus movement of the FocP, followed by wh-movement of the QP sub-extracted from the FocP. A critical step of this derivation is one where the movement of the FocP contains the QP to enable sub-extraction so that the QP may evade a CF violation in a subsequent derivational step. The proposed analysis receives independent support from the amelioration of CF effects under focused wh-questions in Japanese, as reported in Maeda (Reference Maeda2019).
There is no denying that there is some deep-seated computational principle such as CF which requires optimal one-to-one correspondence between an XP and its scope-discourse function. Suggestions of this sort have also been made by Abels's Interpret Once under Sharing (Reference Abels2012: 13), which states that every syntactic object has only one dedicated scope/discourse position relative to a given criterial feature. I have shown that there are certain cases where this otherwise strict correspondence breaks down. One overall picture that emerges from the current investigation then is that such cases have in common that an XP endowed with a criterial feature properly contains a YP endowed with another criterial feature, with the former having the potential to protect the latter from incurring a CF violation. My hypothesis is that this step affords a leeway for multi-criterial configurations to be satisfied without causing any CF violation.
There are important challenges ahead in the research program developed here, among which is the question of what it is about partial wh-movement in Indonesian that makes sub-extraction of the QP alone from within the FocP possible. A quick survey of the literature on CF effects (given in Table 1) suggests that CF is cross-linguistically robust, indicating that this derivation must be severely constrained. Note that the works cited in Table 1 are concerned with CF effects brought about by movement of the same XP to more than one criterial position instead of sub-extraction of YP out of the XP. If there were no constraints on when sub-extraction can be employed, then the relevance of CF would be significantly weakened as a general computational principle, an undesirable outcome.
A Typological Survey of CF Effects
The challenge above is linked to the biggest puzzle concerning the very phenomenon of partial wh-movement, namely, why this construction is impossible in many other languages such as English. I hypothesize, following the insights of Cheng (Reference Cheng1991, Reference Cheng, Liu and Takeda1997, Reference Cheng, Lutz, Müller and Stechow2000) and Watanabe (Reference Watanabe1991, Reference Watanabe1992, Reference Watanabe, Baltin and Collins2001) (see also Kuroda Reference Kuroda1965 and Nishigauchi Reference Nishigauchi1990), that the required sub-extraction of the QP is made possible by a particular morphological structure of wh-words in Indonesian such that the interrogative Q operator is morphologically dissociated from the core which provides the wh-word itself. Indeed, this morphological decomposition of a wh-word into the Q operator and the wh-core is independently supported by Cole and Hermon's (Reference Cole and Hermon1998) observation that wh-words in Indonesian/Malay can be used as variables bound by other operators different from the wh-operator, as shown in (47–48).
In (47a) and (48a), the wh-word is bound by the existential quantifier created by reduplicating the question word. In a similar vein, in (47b) and (48b), the wh-word is bound by the existential quantifier realized by the word -pun ‘also’.Footnote 13
The morphological separability characterizing Indonesian wh-words also provides a principled answer to the question raised by a reviewer as to why the wh-phrase is pronounced in the intermediate position (i.e., the tail of the wh-chain created by further movement of the QP to the matrix [Spec, CP]). This question is important in view of the now-standard guiding assumption in the minimalist framework that the so-called overt vs. covert movement distinction is dispensed with in favour of the single-output model of movement (Bobaljik Reference Bobaljik1995, Brody Reference Brody1995, Groat and O'Neil Reference Groat, O'Neil, Abraham, Epstein, Thráinsoon and Jan-Wouter Zwart1996, Pesetsky Reference Pesetsky, Barbosa, Fox, Hagstrom, McGinnis and Pesetsky1998, Nissenbaum Reference Nissenbaum2000) whereby so-called ‘overt’ and ‘covert’ movements are obtained when the phonological component selects the head and tail of a movement chain for pronunciation/externalization, respectively. The forced stranding and pronunciation of a wh-phrase in an intermediate position in a partial wh-movement construction is a simple consequence of its derivation, in which it is the ‘invisible’ Q-operator that undergoes movement to the matrix [Spec, CP] instead of the whole QP that contains the operator and the wh-core. Note that this derivation is still able to capture Saddy's (Reference Saddy, Cheng and Demirdache1991) core observation introduced in section 2 that ‘covert’ wh-movement of a wh-phrase from its intermediate pronounced site to the matrix C-system obeys island effects and yields its wide scope over a universally quantified subject in the matrix clause.
L. L.-S. Cheng (Reference Cheng, Liu and Takeda1997, Reference Cheng, Lutz, Müller and Stechow2000) argues that partial wh-movement is made available in German for the same morphological reason as Indonesian: the paradigm in (49) from this language shows that indefinite expressions are created by combining their corresponding wh-cores with the prefix irgend.
Recall, however, that I have argued that partial wh-movement in Indonesian has a different derivation from its German counterpart not only with respect to the presence/absence of the overt scope marker in the matrix CP, but also with respect to the scope interaction between the partially moved wh-phrase and a universal quantifier in the matrix clause. L. L.-S. Cheng (Reference Cheng, Liu and Takeda1997, Reference Cheng, Lutz, Müller and Stechow2000) analyzes German partial wh-movement in terms of overt wh-feature movement from a wh-phrase in an intermediate position to the matrix C-system. Updating her analysis in line with contemporary minimalist framework, which does away with Chomsky's (Reference Chomsky1995) feature-movement theory, the construction may be reanalyzed as the result of agreement between the matrix C head and the intermediate wh-phrase, thereby also accounting for the impossibility of the aforementioned scope interaction.
The other challenge relates to the more general issue of whether wh-words have the focus-/wh-structure in most, perhaps all, of the languages of the world. If they do, we should then ask, why shouldn't we find Indonesian-style partial wh-movement more frequently in more languages of the world than we do? There is nothing to prevent a wh-phrase in any language from being associated with the two-layered multi-criterial configuration. At the semantic/discourse level, at least, both focus- and wh-phrases contribute new information and create a structured representation consisting of a focus and a presupposition (see Jackendoff Reference Jackendoff1972; Chomsky Reference Chomsky1976; Rochemont Reference Rochemont1978, Reference Rochemont1986; Culicover and Rochemont Reference Culicover and Rochemont1983; Horvath Reference Horvath1986; Bresnan and Mchombo Reference Bresnan and Mchombo1987; and Kiparsky Reference Kiparsky, Battye and Roberts1995, among many others). We have also seen that the focus feature is morphologically manifested in wh-phrases in Japanese, as in nani-sae ‘what-sae’. Adopting the view that wh-movement instantiates focus movement, suppose then, following a version of Sabel's (Reference Sabel, Lutz, Müller and Stechow2000) theory of wh-movement, that if the [+wh] feature is realized in the matrix C, a [+focus] feature always co-occurs in all the lower embedded C heads, thereby triggering successive-cyclic focus movement, as schematically depicted in (50).
(50)
In this system, whether partial wh-movement is obtained in a language depends on the availability of the stranding of the wh-phrase in the intermediate CPs. This option, I hypothesize, is linked to the morphological decomposability of a wh-phrase into the interrogative operator and the wh-core. Languages such as Indonesian, Ancash Quechua (Cole Reference Cole1982), German (McDaniel Reference McDaniel1989), and Kikuyu (Sabel Reference Sabel, Lutz, Müller and Stechow2000) allow a wh-phrase to stop halfway thanks to their transparent wh-composition system, which makes it possible for the abstract wh-operator/wh-feature to undergo further movement to the matrix [Spec, CP] independently of its containing focused phrase. By contrast, many other languages, including English, do not allow this structure due to their opaque/fusional wh-paradigms, even though their wh-structure is in principle associated with the two-layered feature structure. This property will consequently force movement of the entire focus-/wh-complex, yielding the full wh-movement pattern to the matrix CP. The Serbo-Croatian case investigated in Bošković (Reference Bošković2008) (see Table 1) in which focus movement cannot feed wh-movement can be analyzed along these lines: Since sub-extraction of the QP from the FocP is blocked in this language by an independent language-particular factor having to do with the morphological make-up of wh-words, the entire FocP must undergo focus movement and wh-movement, thereby resulting in a CF violation, unlike in the case of partial wh-movement in Indonesian, even though both constructions seemingly involve the identical sequence of two criterial movements. Nevertheless, the above is a mere speculation at this point, and I will leave a further exploration of the possible link between morphological transparency of wh-words and partial wh-movement for future research.
