The Oxford Research Encyclopedia of Linguistics will be available via subscription on April 26. Visit About to learn more, meet the editorial board, or recommend to your librarian.

Show Summary Details

Page of

PRINTED FROM the OXFORD RESEARCH ENCYCLOPEDIA, LINGUISTICS ( (c) Oxford University Press USA, 2016. All Rights Reserved. Personal use only; commercial use is strictly prohibited. Please see applicable Privacy Policy and Legal Notice (for details see Privacy Policy).

date: 22 March 2018

Scrambling in Korean Syntax

Summary and Keywords

Scrambling is one of the most widely discussed and prominent factors affecting word order variation in Korean. Scrambling in Korean exhibits various syntactic and semantic properties that cannot be subsumed under the standard A/A'-movement. Clause-external scrambling as well as clause-internal scrambling in Korean show mixed A/A'-effects in a range of tests such as anaphor binding, weak crossover, Condition C, negative polarity item licensing, wh-licensing, and scopal interpretation. VP-internal scrambling, by contrast, is known to be lack of reconstruction effects conforming to the claim that short scrambling is A-movement. Clausal scrambling, on the other hand, shows total reconstructions effects, unlike phrasal scrambling. The diverse properties of Korean scrambling have received extensive attention in the literature. Some studies argue that scrambling is a type of feature-driven A-movement with special reconstruction effects. Others argue that scrambling can be A-movement or A'-movement depending on the landing site. Yet others claim that scrambling is not standard A/A'-movement, but must be treated as cost-free movement with optional reconstruction effects. Each approach, however, faces non-trivial empirical and theoretical challenges, and further study is needed to understand the complex nature of scrambling. As the theory develops in the Minimalist Program, a variety of proposals have also been advanced to capture properties of scrambling without resorting to A/A'-distinctions.

Scrambling in Korean applies optionally but not randomly. It may be blocked due to various factors in syntax and its interfaces in the grammar. At the syntax proper, scrambling obeys general constraints on movement (e.g., island conditions, left branch condition, coordinate structure condition, proper binding condition, ban on string vacuous movement). Various semantic and pragmatic factors (e.g., specificity, presuppositionality, topic, focus) also play a crucial role in acceptability of sentences with scrambling. Moreover, current studies show that certain instances of scrambling are filtered out at the interface due to cyclic Spell-out and linearization, which strengthens the claim that scrambling is not a free option. Data from Korean pose important challenges against base-generation approaches to scrambling, and lend further credence to the view that scrambling is an instance of movement. The exact nature of scrambling in Korean—whether it is cost-free or feature-driven—must be further investigated in future research, however. The research on Korean scrambling leads us to the pursuit of a general theory, which covers obligatory A/A'-movement as well as optional displacement with mixed semantic effects in languages with free word order.

Keywords: Korean, scrambling, A-movement, A'-movement, reconstruction, binding, islands, cyclic linearization, free word order

1. Scrambling in Korean

Korean exhibits a wide range of flexibility in word order and provides a rich set of data that can be used to investigate word order variations in depth. As described in (1a), the canonical ordering of a transitive clause in Korean is Subject-Object-Verb (SOV). The object, however, may also precede the subject (OSV), as in (1b). Since Ross (1967), scrambling has been employed as a cover term to describe optional order variations such as (1b). In Korean, not only the object but also other phrases such as the subject, the indirect object, clausal arguments, and certain types of adjuncts may be scrambled to non-canonical position, and more than one constituent may be scrambled in a sentence. Scrambling in Korean is possible across a clausal boundary as well as within a clause.


Scrambling in Korean Syntax

The purpose of this article is to provide a brief introduction and review of word order variations in Korean, with special focus on leftward scrambling. In particular, three issues are examined: (i) What are the main characteristics of scrambling in Korean? (ii) Which factors constrain scrambling in Korean? (iii) What makes scrambling possible in the grammar? Our discussion will be based on data from Korean, but reference to major works on other scrambling languages (e.g., German, Hindi, Japanese) will also be made in order to understand the theoretical import of Korean data from general linguistic perspectives.1

2. Different Types of Scrambling in Korean

Scrambling can be divided into three types, depending on the length of dependency: (i) clause-internal scrambling, (ii) clause-external scrambling, and (iii) VP-internal scrambling. Section 2 examines defining properties of each type of scrambling in Korean. Some noteworthy differences between phrasal and clausal scrambling are also discussed in Section 2. Most previous studies reviewed in this section assume that scrambling is a type of movement. For ease of presentation, the same assumption is adopted in this article (but see section 4 for different approaches to formal properties of scrambling).

2.1 Scrambling and A/A'-Diagnostics

It has been widely accepted in the literature that phrasal movement is categorized into two subtypes: A-movement and A'-movement (Chomsky, 1981). A-movement typically targets the external subject position, SpecTP (e.g., passivization, subject-to-subject raising). A'-movement targets non-theta positions such as SpecCP and adjoined positions (e.g., wh- movement, topicalization). Each type of movement exhibits a range of different syntactic properties. The English examples in (2) and (3) illustrate this point. As shown in (2a), A-movement of John and Mary may establish a new binding relationship with each other. By contrast, A'-movement of who cannot create such a new binding relation, as demonstrated in (3a). The contrast between (2b) and (3b) shows that the pronominal his can be bound by the A-moved phrase everyone, but that such binding is impossible by the A'-moved phrase who. The ill-formedness of (3b) is known as Weak crossover (WCO: Postal, 1971). The examples in (2c) and (3c) illustrate that A'-movement obligatorily reconstructs, whereas A-movement is not required to do so (see Chomsky, 1995; Fox, 1999; Lasnik, 1999; Takahashi & Hulsey, 2009). John in (2c) may be interpreted at the A-moved position, but John in (3c) must be reconstructed to its base position after A'-movement, which results in Condition C violation.2 Other diagnostics such as parasitic gap licensing and quantifier stranding are also employed to make the distinction of A- vs. A'-movement (see Richards, 2014 for an overview).


Scrambling in Korean Syntax


Scrambling in Korean Syntax

A number of earlier studies (especially from the 1980s to the late 1990s) attempted to show that scrambling can be characterized on the basis of A/A'-distinction or by certain modifications of it (Saito, 1985, 1989, 1992; Webelhuth, 1989; Tada, 1993; Mahajan, 1990; Miyagawa, 1997, 2001; Karimi, 2005; see papers in Corver & van Riemsdijk, 1994; Karimi, 2003; Sabel & Saito, 2005 for perspectives on scrambling). Korean scrambling adds some interesting and challenging data to this discussion (Cho, 1994a, 1994b, 1996; Y. Lee, 1993, 1994; Y.-H. Kim, 1999; Jung, 2002; R. Kim, 2003; K. Lee, 2003; Lee & Cho, 2003a; Y. Choi, 2004a; H. Lee, 2006). A general consensus is that scrambling in Korean cannot be mapped into the “standard” dichotomy of A/A'-movement, and that a careful study is yet to be done to claim such a mapping.3

2.1.1 Clause-Internal Scrambling

For ease of presentation, the effects expected from A-movement are called A-effects, and the effects from A'-movement, A'-effects. The examples in (4)–(6) suggest that clause-internal scrambling in Korean exhibits A-effects, just like A-movement in English (2). As described in (4), the anaphor selo ‘each other’ can be licensed by scrambling of kutul-ul ‘they-Acc,’ as seen with anaphor binding in (2a). The wh-phrase, nwukwu-lul in (5), may undergo scrambling to the left of a coreferential pronoun, ku-uy apeci-ka ‘his father-Nom.’ The grammaticality of (5) indicates that clause-internal wh-scrambling does not yield WCO effects, similar to A-movement shown in (2b). Moreover, it also demonstrates that wh-scrambling in Korean significantly differs from the wh-movement in English: the former is not sensitive to WCO effects, whereas the latter is. In (6), we also observe that scrambling of ku-lul ‘he-Acc’ creates a new binding relationship. Minswu in (6) is bound by the scrambled ku-lul and (6) is thus ruled out by Condition C. The evidence adduced in (4)–(6) thus seems to suggest that reconstruction does not occur after clause-internal scrambling in Korean, similar to A-movement in English.


Scrambling in Korean Syntax


Scrambling in Korean Syntax


Scrambling in Korean Syntax

Puzzlingly enough, however, scrambling in (7)–(10) exhibits the opposite pattern, showing A'-effects. As described in (7), the scrambled anaphor caki ‘self’ can be licensed by ku-ka ‘he-Nom.’ This means that the scrambled caki may be interpreted in its original position despite scrambling. The Condition C violation in (8) further shows that Minho-uy emma-lul ‘Minho’s mother’ must be interpreted in its base position. If Minho could be interpreted in the scrambled position higher than ku-ka ‘he-Nom,’ (8) would be acceptable, contrary to fact. Note that the ungrammaticality of (8) is in sharp contrast with English (2c), where A-movement bleeds the Condition C violation. Taken together, the data in (7)–(8) seem to indicate that scrambling does not create a new binding relationship like the A'-movement seen in (3). Ironically, this conclusion is the exact opposite from the one we drew from data in (4)–(6).4


Scrambling in Korean Syntax


Scrambling in Korean Syntax

Moreover, (9)–(10) illustrate that clause-internal scrambling over a (non-contrastive/theme) topic phrase shows consistent A'-effects (Cho, 1994a). In (9), the object John-ul has undergone scrambling over the topic phrase caki-uy sensayngnim-un ‘self’s teacher-Top.’ Interestingly, John in (9) cannot create new A-binding with respect to caki, and the sentence remains ungrammatical even after scrambling. This sharply contrasts with the grammaticality of (4), where scrambling feeds A-binding. In (10), the wh-phrase nwukwu-lul has undergone scrambling to the left of the topic phrase, ku-uy atul-un ‘his son-Top.’ Notably, this type of scrambling yields WCO effects, in contrast to (5). The ungrammaticality of (9)–(10) strongly suggests that scrambling over a topic phrase must be treated differently from the cases exemplified in (4)–(5) (cf. Cho, 1994a, pp. 122–135 for contrastive topics).


Scrambling in Korean Syntax


Scrambling in Korean Syntax

2.1.2 Clause-External Scrambling

It has been argued for Hindi and Japanese that clause-external scrambling shows A'-effect, whereas clause-internal scrambling may show A or A'-effects (Mahajan, 1990 for Hindi; Saito, 1992; Tada, 1993 for Japanese; cf. Saito, 1992, p. 109 for a contrast between Japanese and Hindi in WCO).5 Interestingly, however, it has been reported that Korean shows mixed A/A'-effects not only for clause-internal scrambling but also for clause-external scrambling.

Consider first (11)–(14), which conform to the A'-pattern reported in other languages. The scrambled phrases in (11)–(14) are interpreted in base-position with respect to anaphor binding, Condition C, wh-scope, and NPI licensing. In (11), the anaphor caki is licensed though its surface position is higher than its licenser ku-ka ‘he-Nom.’ This indicates that caki-uy atul ‘self’s son’ can be licensed in its base-position. In (12), John has undergone scrambling over ku-ka ‘he-Nom,’ where it could obviate the Condition C violation. Importantly, however, (12) is ungrammatical. This indicates that John-uy atul-ul ‘John’s son’ must be interpreted in its base-position, violating Condition C.


Scrambling in Korean Syntax


Scrambling in Korean Syntax

In (13), the wh-phrase mwues-ul has undergone scrambling over the matrix subject, but it takes scope in the embedded clause and is licensed by the question morpheme nunci (as an instance of radical reconstruction in the sense of Saito, 1989). In Korean, a Negative Polarity Item (NPI), amwuto ‘anyone’ must be licensed by a clause-mate negation (see Sohn, 1995; Sells, 2015 for overviews). The scrambled amwuto in (14), however, can be licensed by the negation in the embedded clause. This suggests that amwuto in (14) may be licensed after reconstruction at LF (cf. note 6 for some qualification).


Scrambling in Korean Syntax


Scrambling in Korean Syntax

Given the data in (11)–(14), clause-external scrambling in Korean seems to be ordinary A'-movement, which undergoes reconstruction at LF. Interestingly, however, (15)–(18) illustrate the opposite point. In (15), the pronoun kutul-ul ‘they-Acc’ has undergone clause-external scrambling, and it may license the anaphor selo. Though there is some controversy about the grammatical status of (15), this type of example has been judged acceptable by not a few speakers of Korean (Cho, 1994a, 1994b; K. Lee, 2003; H. Lee, 2006; contra Mahajan, 1990 for Hindi and Saito, 1992; Tada, 1993 for Japanese). This means that clause-external scrambling may create a new binder, unlike typical A'-movement. The data in (16) also shows that Condition C is evaluated at the scrambled position, where ku c-commands the matrix subject.


Scrambling in Korean Syntax


Scrambling in Korean Syntax

The pattern observed in NPI scrambling and wh-scrambling also indicates the same point as the binding data in (15)–(16). In (17), the scrambled amwukesto can be licensed by the matrix negation. The example in (18) shows that a scrambled wh-phrase can be interpreted in its surface position. Though it is controversial, researchers often report that scrambled wh-phrases as in (18) can be interpreted either in base-position or in scrambled position (Y. Lee, 1993; Kang & Müller, 1996; Johnston & Park, 2001; J-.M. Yoon, 2013; Jung, 2015; cf. Takahashi, 1993 for Japanese). If clause-external scrambling implicated obligatory reconstruction at LF, the facts in (15)–(18) would not be expected.6


Scrambling in Korean Syntax


Scrambling in Korean Syntax

Moreover, clause-external scrambling of quantifiers in Korean strongly affects semantic interpretation. In (19a), the scrambled quantifier, manhun salam-ul ‘many people’ must be interpreted as a specific set of people, in contrast to (19b) without scrambling (Sohn, 1995, p. 199). If we assume that manhun salam-ul must undergo reconstruction at LF and be interpreted as such, the asymmetry between (19a) and (19b) is not expected (cf. Tada, 1993).7 The example in (20) also hints that clause-external scrambling in Korean is not typical A'-movement. As described in (20), scrambling of nwukwu-lul ‘who’ to the left of the matrix subject does not trigger the WCO effect, in contrast to wh-movement in English (3b) (Cho, 1994a, 1994b; cf. Y. Choi (2004a, p. 188) for an opposing view). Clearly, these facts suggest that clause-external scrambling in Korean is neither semantically vacuous nor a typical type of A'-movement (see Jung, 2002; R. Kim, 2003; K. Lee, 2003; H. Lee, 2006 for further discussion).


Scrambling in Korean Syntax


Scrambling in Korean Syntax

2.1.3 VP-Internal Scrambling

Though it is rather controversial what the base order in double object constructions in Korean is (see J. Lee, 2004; L. Kim, 2015 for overall discussions), it is generally accepted that the direct object may scramble over the indirect object, as illustrated in (21)–(24). This is called short scrambling or VP-internal scrambling. Interestingly, unlike other types of scrambling, VP-internal scrambling shows consistent A-effects (Cho, 1994a, 1994b; see Mahajan, 1990 for Hindi; Saito, 1992; Tada, 1993 for Japanese). As in (21), anaphor binding into the indirect object is possible via VP-internal scrambling. Condition C is evaluated in the scrambled position in (22). Note that the grammaticality of (22) is in contrast to the ungrammaticality of (8) and (12), where Condition C is evaluated in the base-position. Anaphor binding is also evaluated after scrambling, as in (23). The example in (24) shows that wh-scrambling over an indirect object does not trigger WCO violations, just like typical A-movement. Capitalizing on these facts, previous studies argued that VP-internal scrambling in Korean targets A-position (e.g., scrambling to SpecVP in Cho, 1994b; scrambling to SpecAgroP in Cho, 1994a, 1996; object shift to SpecvP in Lee & Cho, 2003a; cf. Hoji, 1985; Nemoto, 1993; Takano, 1998; Saito, 2003, among others, for Japanese).8


Scrambling in Korean Syntax


Scrambling in Korean Syntax


Scrambling in Korean Syntax


Scrambling in Korean Syntax

2.1.4 Clausal Scrambling

In Korean, a clausal element may undergo scrambling to the left of the matrix subject, as in (25c) and (26c). Interestingly, CP scrambling in Korean exhibits consistent reconstruction effects, in contrast to DP scrambling. Kwon (2010), in particular, shows that an anaphor in a scrambled clause does not create any new binding relation, in contrast to DP scrambling, as shown by the contrast between (25b) and (25c). The data in (26) illustrate this with respect to WCO effects. Clause-external scrambling of a wh-phrase does not trigger a WCO violation, as in (26b) (recall (20)). By contrast, scrambling of a CP which embeds nwukwu-lul is unacceptable, as in (26c). This means that the wh-phrase in (26c) is interpreted in its original position together with the embedding clause, just as in (26a). Kwon (2010) provides further evidence for obligatory reconstruction effects of CP-scrambling from NPI licensing and scope data, and claims that CP-scrambling in Korean is semantically vacuous PF-movement.


Scrambling in Korean Syntax


Scrambling in Korean Syntax

2.2 Approaches to Korean Scrambling and A/A'-Movement

Overall, the data examined in Section 2.1 show that scrambling in Korean exhibits non-uniform behavior with respect to A/A'-diagnostics. The obvious question is how to explain these facts with a general theory of scrambling. This section critically reviews three approaches to scrambling in Korean, which naturally connect to the current research agenda in the Minimalism Program (Chomsky, 1995 and subsequent works).

The first approach is to view scrambling as A-movement with special reconstruction effects in certain contexts. Y. Lee (1993, 1994) develops a comprehensive theory of scrambling based on this approach. Y. Lee argues that scrambling in Korean is uniformly Case-driven A-movement (cf. Mahajan, 1990; Miyagawa, 1997, 2001 for A-scrambling). Under Y. Lee’s approach, A-effects obtained with scrambling naturally follow from the claim that scrambling is a kind of A-movement (via IP-adjunction). However, the A'-effects seen in Section 2.1 require additional explanations. Y. Lee claims that scrambling may show A'-effects when it occurs across the subject. In those cases, scrambled phrases must be reconstructed in order to restore a predicational structure at LF, where the subject c-commands all other arguments. For example, A'-effects seen in (7)–(9) and in (11)–(14) may be accommodated by this claim.

On a closer examination, however, it is not clear whether Y. Lee’s proposal may capture all the mixed characteristics of scrambling. Y. Lee assumes that the phrases scrambled over the subject must be reconstructed at LF. If reconstruction were only optional, we would expect (8) to be grammatical, contrary to fact. Crucially, however, this assumption incorrectly nullifies A-effects. For instance, if reconstruction below the subject were obligatory, Condition C would be wrongly obviated in (6). Anti-reconstruction effects observed in (15)–(18) would raise the same problem. Under Y. Lee’s approach, it is expected that scrambling will not trigger WCO effects because the target position of scrambling is Case-position with A-properties. Thus, the contrasts between (5) and (10) or between (10) and (20) would not be explained by Y. Lee’s theory in any straightforward way. More generally, one may question Y. Lee’s premise that scrambling is Case-driven. It is unclear how this may be compatible with the fact that PPs, adjuncts, and clauses can be scrambled in Korean which do not require Case in syntax. Moreover, it is not yet obvious in what sense clause-external scrambling is Case-driven, given that a scrambled phrase receives Case clause-internally (cf. responses by Lee, 1993, pp. 122–137).9

The second approach is to assume that scrambling can indeed be A- or A'-movement, depending on the landing site (Mahajan, 1990; Saito, 1992; cf. Saito, 2003). This is an approach extensively developed by Cho (1994a, 1994b, 1996). Contra Y. Lee (1993), Cho argues that scrambling in Korean is not a uniform phenomenon, but must be divided into three subcategories. The first type is short scrambling to SpecAgroP, illustrated in (21)–(24). Cho argues that short scrambling shows consistent A-effects (anti-reconstruction effects) and must be treated as A-movement. Cho argues that the second type is non-operator A'-scrambling (following Webelhuth, 1989 and Saito, 1992).10 Clause-internal scrambling over a subject (e.g., (4)–(8)) and clause-external scrambling (e.g., (11)–(18)) are categorized into this type. Cho argues that this type of scrambling targets IP adjunction or VP adjunction positions. In this case, scrambling shows A'-effects because it targets an adjunction position, and at the same time, it obviates WCO effects because the landing site is a non-operator position and creates a null epithet (cf. Lasnik & Stowell, 1991). The last type is operator A'-scrambling over a topic phrase. The examples in (9)–(10) represent this category. This type of scrambling show coherent A'-effects because its target is an operator A'-position, TopP adjunction (cf. Cho, 1994a, pp. 122–135 for contrastive topics as IP-adjunction; cf. Cho & Kim, 2000; Y. Choi, 2004a for alternative approaches).

Cho’s proposal captures diverse syntactic properties of scrambling by adopting a hybrid approach. It still remains puzzling, however, why mixed A/A'-effects are obtained with a single type of scrambling. In particular, anti-reconstruction effects in non-operator A'-scrambling may potentially challenge the proposal. To explain the contrast between (8) and (22) in binding, Cho assumes that A'-scrambling undergoes obligatory reconstruction at LF (cf. Cho, 1994a, pp. 65–70 for WCO effects). This assumption, however, leads us to wrongly predict that R-expressions may avoid Condition C violations after reconstruction, contrary to the facts in (6) and (16). It remains unexplained how this assumption can be made compatible with the fact that reconstruction is only optional for NPI licensing as in (17) or wh-licensing as in (18). The status of anaphor binding also requires further explanation. To explain the contrast between (4) and (9), Cho (1994a, 1994b) proposes that anaphors must be non-operator bound (following Saito, 1992). It is unclear, however, when anaphor binding happens. Examples like (7) and (11) suggest that anaphor binding occurs after reconstruction, but (4) and (15) hint that it is evaluated before reconstruction at LF. The general interaction between binding and reconstruction needs to be further clarified (cf. Cho, 1996; Cho & Kim, 2000 for discussion).

Thirdly, one may attribute mixed A/A'-effects of scrambling to optionality of reconstruction at LF. R. Kim (2003) takes this approach. R. Kim argues that scrambled phrases may in principle undergo radical reconstruction at LF, but if necessary for feature checking, they may stay in a scrambled position or in an intermediate position. On this approach, A'-effects in (7)–(9) and in (11)–(14) are observed because scrambling can be freely undone at LF (Saito, 1989). If, however, scrambling is associated with feature checking, it may stay in non-base position where the relevant feature can be checked. R. Kim claims that Case-checking in (4), NPI-licensing in (17), and wh-licensing in (18) instantiate such cases where feature checking overrides radical reconstruction effects at LF (cf. Son, 2001; Jung, 2002; Yang & Kim, 2005; H. Lee, 2006 for alternative views on feature checking and reconstruction in scrambling).

Note, however, that optional reconstruction cancels out A'-effects incorrectly. If the object in (4) may stay in scrambled position for Case checking, we expect that the same would be true of scrambled R-expressions. Such an option, however, would wrongly rule in ungrammatical sentences such as (8) and (12). Moreover, if radical reconstruction is in principle a possible option, we expect that examples like (16) might be saved via reconstruction after feature checking. Recall that in the case of short scrambling such as (21)–(24), LF-reconstruction is not an option, but must be banned. It is not obvious how this approach would accommodate the differences between short scrambling and other types of scrambling. Mixed WCO effects are not expected under this approach, either.

In short, previous studies have discovered major characteristics of scrambling by comparing it with A/A'-movement. The ample empirical discoveries of the previous studies should be well taken, but it also needs to be asked whether the notion of A/A'-distinction is indeed necessary to understand scrambling. As the theory of Minimalist syntax (Chomsky, 1995) develops, the distinction of the A vs. A'-positions becomes nothing but a descriptive notion. In fact, a variety of proposals have been advanced to capture syntactic and semantic properties of scrambling without resorting to A/A'-distinctions (e.g., Abe, 1993; Bošković & Takahashi, 1998; Cho & Kim, 2000; Son, 2001; Jung, 2002; Saito, 2003). Even if the A/A'-distinction is still a useful tool to describe the typology of movement, it is far from clear whether there exist any “standard” A/A'-diagnostics. Some previous studies tacitly assumed that A/A'-diagnostics employed for other languages can be applied to Korean, but it is yet to be seen whether independent justification for this assumption can be provided.

It is well known that anaphors in Korean can be bound by an antecedent across a clausal boundary in certain contexts and behave much differently from local anaphors in other languages (Yang, 1983, 1986; see Madigan, 2015 for an overview).11 Thus, one might reasonably suppose that anaphor binding in Korean has little to do with A-diagnostics developed from other languages (Y. Lee, 1993; Y. Choi, 2004a). Earlier studies assumed that all the binding relationships must be evaluated at LF, and thus binding interpretation at non-surface position was taken as evidence for reconstruction at LF. This assumption, however, has been seriously challenged as well. It has been argued that anaphor binding, pronominal binding, and Condition C effects are evaluated at different points of the derivation and thus cannot be treated in the same way (Belletti & Rizzi, 1988; Barss, 1986; Lebeaux, 1988, 2009; Epstein et al., 1998; Kitahara, 2002; Saito, 2003; Y. Choi, 2004a). On this view, various types of binding data can be duly examined to detect potential semantic effects of scrambling, but may not be taken as a diagnostic to claim LF-reconstruction or A'-effects in general.

Whether we adopt a hybrid approach or pursue a uniform theory of scrambling, it is important to cast Korean data within a general perspective on word order variation. Korean poses empirical challenges to the claim that scrambling can be freely undone at LF (cf. Saito, 1989; Bošković & Takahashi, 1998). Clause-external scrambling in Korean is semantically effective in terms of binding and scope facts, contrary to what was argued for Hindi and Japanese (cf. Mahajan, 1990; Saito, 1992; Bošković & Takahashi, 1998). Scrambling in Korean does not trigger WCO effects even when it occurs across a clausal boundary, unlike Hindi (cf. Mahajan, 1990). It may be the case that A/A'-diagnostics have different implications for each language, or that there are truly different types of scrambling in languages. To respond to these questions, it is necessary to compare scrambling in different languages with a formal theory that can be applied to language in general. How to formulate such a general theory of scrambling remains an important research agenda.

3. Scrambling Is Not a Free Option

Scrambling in Korean is an optional operation, yielding flexible orderings. This, however, does not mean that word order in Korean is randomly determined. In fact, scrambling in Korean is systematically regulated by various factors in the grammar. Section 3 examines major factors that constrain scrambling at the syntax proper and its interfaces with phonology, semantics, and discourse.

First consider the constraints that operate at the syntax proper. It has been reported that scrambling in Korean cannot occur out of island domains. More specifically, scrambling cannot occur across strong islands such as relative clauses and adjunct clauses, as shown in (27) (Y. Lee, 1993, pp. 140–163; Cho, 1994a, pp. 106/131; Y. Choi, 2004a; R. Kim, 2003).


Scrambling in Korean Syntax

Moreover, scrambling cannot occur from the left branch of a noun phrase (Left Branch Condition: Ross, 1967), as shown in (28a). It is also impossible to strand a genitive-marked adnominal modifier via leftward scrambling of the host noun, as in (28b) (cf. Ko, 2014b, 2016 for such possibilities in rightward dislocation in Korean). As illustrated in (29), scrambling out of a coordinated structure is impossible (Coordinate Structure Constraint: Ross, 1967). Scrambling also obeys the Proper Binding Condition (Fiengo, 1977) in that a scrambled phrase cannot contain an unbound trace, as in (30) (see Saito, 1985, 1992 for Japanese). Overall, the facts in (27)–(30) were taken as counterevidence against the claim that scrambled orders are freely base-generated (cf. Hale, 1982; Bošković & Takahashi, 1998 for base-generation approaches). Rather, scrambling in Korean is an instance of a movement operation, which is regulated by general constraints on movement.


Scrambling in Korean Syntax


Scrambling in Korean Syntax


Scrambling in Korean Syntax

A note is in order, however, on islandhood in Korean. Some researchers report that scrambling out of certain types of islands is considerably acceptable in Korean, in contrast to regular A'-movement, such as wh-movement and topicalization in English (Y. Lee, 1993; R. Kim, 2003). Y. Lee (1993) claims that scrambling out of a NP-complement, as in (31a), is possible, in contrast to extraction out of a relative clause seen in (27a). R. Kim (2003) reports that scrambling out of a wh-island is possible in Korean, as in (31b) (see also Y. Lee, 1993, p. 153 for the same point).


Scrambling in Korean Syntax

Y. Lee (1993, p. 163) proposes that subcategorized clauses (e.g., complements) do not constitute an island in Korean, whereas the non-subcategorized (e.g., adjunct and relative) clauses constitute a strong island. On this view, (31a) and (31b) are grammatical, because scrambling occurs out of a complement clause. R. Kim (2003) analyzes wh-islands such as (31b) as weak islands (in the sense of Rizzi, 1990), which must be distinguished from strong islands. R. Kim claims that strong islands constrain all sorts of movement regardless of its type, whereas weak islands block movement for A'-feature checking only. R. Kim proposes that scrambling is not a feature-driven movement, and thus that scrambling is insensitive to weak islands as in (31b) although it is sensitive to strong islands, as in (27a) and (27b). It is yet to be shown, however, how R. Kim’s account may accommodate Y. Lee’s observation concerning lack of island effects in NP-complement domains (see Y. Lee, 1993, Chapters 5–6 for further discussion concerning complexity of islandhood in Korean).

It has been observed that there is an argument-adjunct asymmetry in the length of scrambling. Though clause-external scrambling of an argument is readily available in Korean (Section 2.1), clause-external scrambling of an adjunct is severely limited (Cho & Kim, 2000; cf. H. Lee, 2006, pp. 453–454 for an opposing claim; cf. Saito, 1985; Bošković & Takahashi, 1998 for Japanese). As in (32a), when an adjunct PP undergoes clause-external scrambling, speakers find it difficult to interpret it as a modifier for the embedded clause (but some speakers accept it when the PP is a temporal/locative phrase). As in (32b), clause-external scrambling of an adverb is not possible, either. If acceptable, sikkulepkey ‘loudly’ is interpreted as a modifier for the matrix verb in (32b).


Scrambling in Korean Syntax

Y. Lee (1993) claims that certain types of scrambling may be blocked due to an anti-ambiguity strategy, which is assumed to be a discourse constraint (following Kuno, 1980). As described in (33) and (34), when two arguments are marked by identical Case, scrambling of the lower element over the higher one is unacceptable. As illustrated in (33b), the nominative complement of a stative verb cannot move over the experiencer marked by nominative Case. If the complement is marked by dative Case, as in (33c), such scrambling becomes available. Similarly, the examples in (34) show that a dative-marked argument in an embedded clause cannot be scrambled over a dative-marked matrix argument. If scrambling were just a random option, we would expect that all the sentences in (33)–(34) would be equally acceptable, contrary to fact. Y. Lee (1993, pp. 117–118) argues that speakers tend to assign the base order interpretation when a sentence is potentially ambiguous between scrambled and non-scrambled structure. For instance, (33a) is potentially ambiguous due to having two NPs with identical Case marking. It could mean ‘I am fond of Minho’ with a non-scrambled order, or ‘Minho is fond of me’ with a scrambled order. Though both structures are in principle available in syntax, the scrambled parse is disfavored to avoid ambiguity, and (33a) is interpreted with a non-scrambled structure only.


Scrambling in Korean Syntax


Scrambling in Korean Syntax

It is generally assumed that string vacuous scrambling is banned. An important piece of evidence for this restriction can be drawn from scope rigidity of canonical sentences (see Hoji, 1985 for the original argument based on Japanese). In Korean, the subject scopes over the object in the canonical order as in (35a). If, however, the object undergoes scrambling over the subject, as in (35b), scope ambiguity arises: the object may scope over or under the subject (Ahn, 1990; Suh, 1990; S. Kim, 1991; Sohn, 1995, p. 145; Son, 2001; cf. Kim & Larson, 1989 for scope in psych-predicate constructions; see Huang, 1982 for Chinese, Hoji, 1985 for Japanese). Suh (1990) argues that a quantifier phrase QP1 may scope over QP2 if QP1 c-commands a member of the chain containing QP2 (Aoun & Li, 1989). The scrambled object in (35b) c-commands the subject, which in turn c-commands the trace of the object. Thus, we obtain scope ambiguity in (35b). There is a missing piece in this explanation, however. If the subject in (35a) undergoes string vacuous scrambling over the object, as in (35c), we would expect (35a) to be ambiguous, contrary to fact. To block such a possibility, it is necessary to assume that string vacuous scrambling such as (35c) is somehow ruled out. In the Minimalist Program, the ban on string vacuous scrambling can be attributed to interface economy: optional movement occurs only when it has an effect at the interface (see Chomsky, 1995, 2001; Fox, 2000; Miyagawa, 2006 for output economy). Sabel (2005) claims that scrambling must have an output effect at both PF and LF and that string-vacuous scrambling without a PF-effect such as (35c) is not allowed (see Sabel, 2005 for further discussion).12


Scrambling in Korean Syntax

When two NPs with identical Case compete for scrambling, semantic factors may play a crucial role as well. As illustrated in (36b), when two NPs are semantically associated by a part-whole relationship, the part NP cannot be scrambled over the whole NP. A similar restriction holds on two NPs in a subset-superset relationship, as in (37b). The examples in (38) illustrate an ordering restriction in multiple nominative constructions. The first NP in a multiple nominative construction is called the Major Subject, and the following (saturated) sentence functions as a sentential predicate which denotes a characteristic property of the Major Subject (J. Yoon, 2004, 2007; see Kuno, 1973 for original insight). The Major Subject can be an argument of the verb, a possessor of the grammatical subject, or a scene-setting adjunct such as a locative PP or a source PP (see J. Yoon, 2015 for an overview). Notably, the Major Subject cannot be preceded by other nominative-marked NPs via scrambling, as in (38b) (see J. Yoon, 2004).


Scrambling in Korean Syntax


Scrambling in Korean Syntax


Scrambling in Korean Syntax

The data in (36)–(38) cannot be subsumed under the anti-ambiguity strategy introduced for (33)–(34), because they are not ambiguous. The data in (36)–(38) cannot be explained by output economy, either, because scrambling of the second NP is not string vacuous. Rather, the data in (36)–(38) suggest that certain semantic considerations may regulate scrambling possibilities in constructions with multiple Case marking. In particular, adopting Kuno’s (1973) aboutness condition, J. Yoon (2004) argues that Major Subjects in Korean must be ‘news-worthy’ and claims that the Major Subject must precede the sentential predicate (including the Grammatical Subject) to be interpreted as such (for alternative proposals on this restriction, see Lee & Cho, 2003b for a locality-based approach; Ko, 2014a, pp. 170–175 for a cyclicity-based approach). It remains open, yet, whether the ordering restriction imposed on multiple Accusative constructions shown in (36)–(37) should receive the same account as the one for multiple Nominative constructions.

Case omission also interacts with the (im-)possibility of scrambling. As illustrated in (39a), the accusative Case of the object in Korean can be dropped in the canonical SOV order (see Kwon & Zribi-Hertz, 2008 and H. Lee, 2015, among others, for Case ellipsis in Korean). Interestingly, if an indefinite object is devoid of overt Case, object scrambling is judged degraded, as in (39b). A similar pattern is found in (40). Lee and Cho (2003a, 2003b) argue that an object must undergo object shift prior to scrambling, and that an indefinite object may undergo object shift only when it is overtly Case-marked and interpreted as specific. On this view, the objects in (39b) and (40b) cannot be scrambled, because they are interpreted as non-specific without overt Case (cf. Y.-H. Kim, 1998 for an alternative view).13


Scrambling in Korean Syntax


Scrambling in Korean Syntax

Scrambling may be ruled out due to an interplay between the syntax and its interfaces as well. A series of works by Ko (2005a, 2007, 2011, 2014a) provides an in-depth discussion on this issue. Ko proposes that cyclic Spell-out crucially affects the (re)ordering of elements in scrambling, and shows that certain instances of scrambling are filtered out at the PF interface even though they are perfectly grammatical at the syntax proper. Evidence for this claim is drawn from a wide range of asymmetries obtained in scrambling in Korean and Japanese. Two representative pieces of evidence are introduced in this article: subject scrambling and predicate inversion.

The data in (41) and (42) illustrate a well-known asymmetry between the subject and the object in scrambling. As illustrated in (41), the object in Korean may undergo scrambling over the subject, stranding an associate numeral quantifier (NQ). Interestingly, however, the subject cannot strand an associate NQ sey-myeng over the object, as in (42). If scrambling were just a free option, we would expect that the subject could scramble over the object in (42), contrary to fact. Observing the same type of asymmetry in Japanese, Saito (1985) originally claimed that the subject in general cannot undergo scrambling, and thus examples like (42) are ruled out (cf. Hoji, 1985; Miyagawa, 1989). Ko (2005a, 2007), however, shows that the subject in Korean and Japanese may indeed scramble in some contexts. The subject may scramble over vP-external high adverbs such as ecey ‘yesterday,’ as in (43). Subject scrambling may also occur across a clausal boundary (with some parsing difficulty), as in (44) (cf. Saito, 1985; see Sohn, 1995, p. 236 for parsing difficulties in subject scrambling). The grammaticality of (43) and (44) strongly suggests that the subject can in principle scramble, but subject scrambling is somehow more restricted than object scrambling, as shown by the contrast between (41) and (42) (see Ko, 2005a, 2014a, and references therein for further discussion).


Scrambling in Korean Syntax


Scrambling in Korean Syntax


Scrambling in Korean Syntax


Scrambling in Korean Syntax

Following Fox and Pesetsky (2005), Ko argues that the output of syntax maps into the phonology via Cyclic Linearization (CL), which establishes linear orderings of syntactic terms at each Spell-out (cf. Chomsky, 2000, 2001). Crucially, the linear orderings of a syntactic unit must be preserved once it undergoes CL. Under Ko, it is assumed that scrambling is triggered by a head which contains an EPP feature (or scrambling-related discourse feature), which c-commands its goal (Chomsky, 2001). On this view, scrambling of the object in (41) is possible because the object may undergo scrambling to the left of the subject at vP before Spell-out of the vP domain, as in (45) (see Jung, 2002; Cho, 1994a, 1996; Lee & Cho, 2003a for successive cyclicity in scrambling). Scrambling of the subject in (43) is also possible because the subject may undergo scrambling over vP-external materials when it is probed by a higher head such as T, as illustrated in (46) (see Ko, 2007, pp. 58–59 for detailed descriptions).


Scrambling in Korean Syntax


Scrambling in Korean Syntax

An interesting problem arises when the subject undergoes scrambling over vP-internal material as in (42). As illustrated in (47), the object may optionally undergo scrambling within the vP, and thus two types of linear orderings can be obtained in syntax: (i) S<NQSubj<O or (ii) O<S<NQSubj. Importantly, however, the order in which the object intervenes between the subject and NQSubj cannot be generated. As depicted in (48a), the subject is already merged on the edge of the vP and cannot undergo vP-internal movement (under probe-goal Search: Chomsky, 2001). Thus, S<O<NQSubj order cannot be generated within the vP. If the subject scrambles over the object in a later derivation, the derivation may be licit in syntax but is filtered out at the interface. As illustrated in (48b), if the object intervenes between the subject and NQSubj in CP, the orderings at CP in (48b) necessarily conflict with the possible orderings at vP demonstrated in (47). In short, (42) is ruled out not because subject scrambling is totally banned, but because the word order at vP must be preserved after CL (see Ko, 2014a for further evidence and formal descriptions of CL effects).14


Scrambling in Korean Syntax


Scrambling in Korean Syntax

The examples in (49)–(50) illustrate how CL restricts predicate inversion in Korean. As illustrated in (49), the small clause (SC) predicate kyoswu-lo ‘professor-as’ may be fronted over the main subject. Importantly, however, predicate inversion is not always possible. For instance, predicate inversion of a SC-predicate, ceyca-lo is impossible in (50b). The obvious question is why we observe two different types of behavior of small clause predicates in (49)–(50). Ko (2011, 2014a) argues that the crucial difference between the two cases lies in the semantic nature of the main predicate, which affects the size of linearization domain at the interface.


Scrambling in Korean Syntax


Scrambling in Korean Syntax

The main verb ‘hire’ in (49) takes the object as its complement, and (49a) entails that ‘SNU hired Dr. Lee.’ The epistemic verb ‘consider’ in (50), on the other hand, takes a proposition as its complement, and thus (50a) does not entail that Prof. Kim considered Dr. Lee (see Aarts, 1992 for similar distinctions in English). This means that the complementation structures of (49) and (50) are radically different. As depicted in (51), Lee paksa-lul in (49a) is merged as the object of the main verb, in a different domain from the SC-predicate, kyoswu-lo. By contrast, Lee paksa-lul in (50a) is merged in the same domain as SC-predicate, ceyca-lo, as illustrated in (52). RPs in (51)–(52) represent a small clause domain, which undergoes Spell-out and CL at the interface (see Ko, 2011, 2014a, 2015 for the typology and structure of small clauses in Korean; cf. den Dikken, 2006, 2007 for the syntax and semantics of RPs).


Scrambling in Korean Syntax


Scrambling in Korean Syntax

The contrast between (51) and (52) in their argument structures results in crucial differences in linearization. In the case of (51), the object is linearized in a separate domain from the small clause predicate. Thus, the predicate (more precisely, the small clause RP with PRO) can be scrambled over the object in (49b) without ordering conflicts. By contrast, in (52) the SC-subject must be linearized together with its predicate, given that they are merged in the same predicational domain, RP. Since the SC-subject precedes the SC-predicate in the small clause, this ordering must be preserved after CL. If predicate inversion occurs in a later derivation, as in (50b), ordering conflicts arise between the small clause and CP domains, and thus the derivation is filtered out at the interface. Put generally, the CL-approach to scrambling correctly predicts that SC-predicate inversion is banned once the small clause undergoes CL, and crucially, this restriction holds only in the cases where the SC-subject and SC-predicate are externally merged within the same small clause. Ko (2014a) extensively argues that this account extends to restrictions on predicate inversion out of other types of small clauses in scrambling languages (see Ko, 2014a, 2015 for CL effects in depictives, resultatives, and decomposed VPs).

In this section, it has been shown that scrambling in Korean applies optionally but not randomly. It may be blocked due to various factors in syntax and its interfaces in the grammar. Before closing this section, it is worth mentioning that scrambling is sometimes forced if scrambling is the only way to derive a grammatical output. One such case is the NPI licensing discussed with (17); clause-external scrambling of amwukesto is necessary to license it by the matrix negation (Sohn, 1995; Y. Lee, 1993; R. Kim, 2003). A structure with potential “LF-Intervention effects” is another well-established case (Beck & Kim, 1997; cf. Hoji, 1985; S. Kim, 1991). As illustrated in (53a), the NPI amwuto cannot precede a wh-phrase in Korean (cf. Ko, 2005b for exceptional behavior of way ‘why’). In contrast, if the wh-phrase undergoes scrambling over the NPI, as in (53b), it is interpreted as a licit wh-question. Beck and Kim (1997) argues that wh-scrambling in (53b) is forced to obviate intervention effects at LF. Specifically, when the NPI and negation hierarchically intervene between mwues-ul and Q, as in (53a), it blocks wh-licensing at LF. By contrast, if wh-scrambling occurs, as in (53b), the NPI does not intervene between the wh-phrase and Q at LF, and the sentence is interpreted as a licit wh-question. S.-S. Kim (2002) further argues that focus-bearing elements work as an LF-intervener and thus that wh-scrambling is necessary when a wh-phrase is c-commanded by a focus-bearing element in the overt syntax (cf. Sohn, 1995; Son, 2001; Tomioka, 2007; Y. Choi, 2007, among others, for different approaches).


Scrambling in Korean Syntax

4. Formal Properties of Scrambling

Section 4 provides an overview of major studies on the formal property of scrambling, assessing their implication for Korean data. The term scrambling was coined by Ross (1967), who defined it as a stylistic rule which applies freely in the grammar. Its theoretical import has been adopted in many different ways since then. Though the details differ, major approaches can be summarized as in (54): some authors argue that scrambling is a result of base-generation, whereas others take scrambling as a movement operation with differing assumptions on its nature (cf. Corver & van Riemsdijk, 1994, p. 13).


Scrambling in Korean Syntax

Consider first the base-generation approach, which posits non-configurational flat structures. Hale (1982) provides a theoretical account on free word order based on this approach. Hale, in particular, proposes a typology that partitions languages into two groups: configurational and non-configurational. Configurational languages utilize a hierarchical structure, whereas non-configurational languages take a flat structure such as (55). Under the flat structure, it is assumed that all phrases have a symmetrical relation with the head, so that they are free to occur in any order in syntax. Under this proposal, different orders in scrambling languages are freely base-generated in syntax; no movement is involved in scrambling.


Scrambling in Korean Syntax

If Hale’s proposal is on the right track, scrambling languages should exhibit general properties of non-configurational structures in syntax. Subsequent work, however, presented considerable evidence against this prediction (Saito & Hoji, 1983; Saito, 1985; Hoji, 1985; Choe, 1985; Whitman, 1987). The data from Korean lend further evidence against the non-configurationality approach. As seen in Section 2, clear configurational asymmetries are observed in Korean in terms of binding, scope, and crossover effects. As discussed in Section 3, word order variation in Korean is regulated by various predictable factors at the syntax proper and its interfaces. Scrambling in Korean also affects semantic interpretation in areas such as scope. If scrambled orders can be randomly base-generated, the facts discussed in Sections 2 and 3 would not be expected. All of these data point to the conclusion that scrambling in Korean cannot be attributed to the hypothesis that Korean takes the non-configurational parameter.

An alternative view is to inherit Hale’s insight that scrambled orders are freely base-generated, but assume that scrambling languages are fully configurational. Some researchers indeed pursue this line of research (e.g., Y.-S. Kim, 1997; Neeleman & Reinhart, 1998; Bošković & Takahashi, 1998; Cho & Kim, 2000; Fanselow, 2001; Bošković, 2004; cf. also Kempson & Kiaer, 2010 for a processing-based approach under the Dynamic Syntax model). Mainly based on Japanese, Bošković and Takahashi (1998) argue that scrambled phrases must be base-generated in their surface position (cf. Neeleman & Reinhart, 1998 for flexible merger of the object in Germanic OV languages). Under Bošković and Takahashi (1998), the phrases scrambled clause-externally must undergo lowering at LF in order to check their theta-feature. The phrases scrambled clause-internally, on the other hand, may stay in their surface position if locally theta-marked in IP-adjoined position or undergo LF-lowering to its theta position (cf. Cho & Kim, 2000). This analysis makes optional scrambling consistent with the Last Resort principle (Chomsky, 1995) in that scrambled phrases undergo obligatory (covert) movement for feature-checking at LF.

Crucially, this approach predicts that clause-external scrambling would be semantically vacuous, because scrambled phrases must be lowered to their theta position at LF (Bošković & Takahashi, 1998; Bošković, 2004; cf. Saito, 1989). This prediction, however, is not upheld in Korean scrambling with mixed A/A'-effects. As discussed in Section 2, upstairs interpretation of scrambled phrases is sometimes necessary and sometimes optional in Korean. If LF-lowering for theta checking is an obligatory operation (as enforced by Last Resort), anti-reconstruction effects shown in (15)–(20) would remain surprising. Put generally, cross-linguistic variations in semantic effects of scrambling are not expected under this approach. Moreover, if scrambled phrases are base-generated in surface position, no WCO effects (or their variations) would be expected in scrambling, contrary to fact (recall (5) vs. (10)). More importantly, it remains largely unexplained how this approach would explain the fact that movement constraints posited in the overt syntax regulate scrambling. As previously mentioned, scrambling in Korean is sensitive to movement restrictions such as strong islands, PBC, LBC, and CSC, showing locality effects. This, however, is not be expected under the base-generation approach in an obvious way (cf. Bošković & Takahashi, 1998, pp. 358–359/361 for some responses; see Bailyn, 2001; Johnston & Park, 2001; and Miyagawa, 2006 for extensive reviews and criticisms on the base-generation/LF-lowering approach to scrambling). Moreover, the set of evidence adduced for the claim that scrambling undergoes cyclic Spell-out in the course of derivation cannot be accommodated in any straightforward way under the base-generation approach (see Ko, 2014a).

Against the base-generation approach, a number of researchers argue for a movement approach to scrambling. Some argue that scrambling is an instance of movement that operates at the phonological component, not in core syntax (see Ross, 1967; Chomsky & Lasnik, 1977 for scrambling as a stylistic rule; Zubizarreta, 1998 for prosodically motivated scrambling; and Sauerland & Elbourne, 2002 for PF-scrambling in Japanese). Kwon (2010), in particular, adopts this approach and argues that CP-scrambling in Korean is optional PF-movement driven by prosodic requirement and has no LF import at all. Though this approach may fit well with the CP-scrambling data which show total reconstruction effects, it cannot be extended to other instances of scrambling with semantic effects. As demonstrated in Section 2, certain types of DP-scrambling in Korean clearly yield new semantic interpretations. This means that PF-movement cannot be the sole source of scrambling in Korean. If PF-scrambling exists in the grammar, it is crucial to clarify how PF-scrambling can be dissociated from syntactic scrambling in a principled way and why that is desirable. These questions, however, will have to be answered in future research.

Another approach in this vein is to regard scrambling as cost-free optional movement in syntax—which has been the most dominant view in the field, especially in the pre-minimalist era (e.g., Kuroda, 1988; Saito, 1985, 2003; Hoji, 1985; Fukui, 1993; Abe, 1993; Cho, 1994a, 1994b; Saito & Fukui, 1998; Tada, 1993; Takano, 1998; R. Kim, 2003). In this approach, scrambling is considered a truly optional and costless movement, which may occur without any driving force. Fukui (1993) and Saito and Fukui (1998), in particular, argue that optional movement in language depends on the directionality of the head parameter: left-headed languages allow optional movement to the right, such as heavy NP shift in English, whereas right-headed languages allow optional movement to the left, such as leftward scrambling in Korean and Japanese. This approach naturally captures the fact that scrambling in a head-final language yields an optional variation, unlike obligatory A/A'-movement in other languages (cf. Bailyn, 2001, p. 652 for right-branching scrambling languages such as Russian as a challenge to this theory). Since scrambling is defined as a type of movement, it straightforwardly follows that scrambling is subject to movement constraints, as in (27)–(31), and it is also expected that scrambling may yield a new interpretation in terms of binding and scope, as a consequence of displacement in syntax.

As the theory of syntax develops in the framework of the Minimalist Program, however, this approach faces some non-trivial challenges both empirically and theoretically. On the theoretical side, this approach is not compatible with the premise that syntactic operations are driven by morphological forces such as Last Resort (Chomsky, 1995) or probe-goal Search (Chomsky, 2000, 2001). Under this approach, it remains largely open when and why scrambling occurs in a language, because scrambling is assumed to be a cost-free option. On the empirical side, it cannot explain the fact that scrambling is constrained by the same factors that regulate feature-driven movement. If scrambling is a cost-free option, we expect that scrambling would be much freer than feature-driven movement, but this prediction is not borne out. It has been shown that scrambling exhibits locality and anti-locality effects just as feature-driven movement does (Miyagawa, 2001, 2010; Richards, 2001; Kitahara, 2002; Ko, 2007, 2011, 2014a). Scrambling is allowed only when a proper probe (trigger) exists within the cyclic domain, just as in the case of feature-driven movement (e.g., (48a)).

Some recent studies converge on the conclusion that scrambling is a feature-driven movement in the syntax proper (e.g., Y. Lee, 1993; Miyagawa, 1997, 2001, 2010; Karimi, 1999; Grewendorf & Sabel, 1999; Sabel, 2001; Kitahara, 2002; Lee & Cho, 2003a, 2003b; Jung, 2002; Yang & Kim, 2005; H. Lee, 2006; Ko, 2007, 2011, 2014a). Notably, this approach claims that scrambling is obligatory movement despite the fact that scrambling yields an optional variation. A head may optionally obtain a scrambling feature, but once the feature is assigned, scrambling becomes obligatory in syntax. Proponents of this approach argue that scrambling must be triggered by a formal feature, like other types of movement in syntax, and thus that it is constrained by general principles for feature checking. Given that scrambling is a type of movement, it naturally explains the fact that scrambling shows the traits of movement in its derivation and representation; it is subject to movement constraints and may create a new semantic interpretation, as seen in Sections 2 and 3. This model is in good harmony with current assumptions on cyclic syntax, so it is expected that scrambling will be filtered out if it cannot maintain the monotonicity of CL at PF. It is also correctly predicted that scrambling will not be possible if a proper probe does not exist within the cyclic domain, just as for feature-driven movement (Ko, 2014a for general discussion).

The nature of the formal feature that triggers scrambling is under investigation, however. Various proposals have been advanced: Case (Y. Lee, 1993; cf. Miyagawa, 1997 for A-scrambling), Σ‎-feature (Grewendorf & Sabel, 1999), EPP (Miyagawa, 2001; Kitahara, 2002; cf. INT-Move by Yang & Kim, 2005), Focus/Topic (Karimi, 1999; Miyagawa, 1997, 2010, 2017; Jung, 2002; Lee & Cho, 2003a, 2003b), and Edge Feature (H. Lee, 2006), among many others. It has been widely reported that scrambling in Korean is associated with discourse effects such as topic, focus, specificity, or presuppositionality (Y. Lee, 1993; D. Kim, 1993, 1995; H. Choi, 1999; Lee & Cho, 2003a, 2003b; Son, 2001, 2003). Thus, it seems reasonable to assume that some sort of discourse force underlies the scrambling operation in syntax. It remains open, however, how fine-grained analyses can be developed for the theory of formal features that trigger scrambling. It also remains to be seen how successfully the feature-based theory may accommodate typological variations in scrambling, using the inventory of formal features in languages.

5. Implications of Studies on Korean Scrambling

This article has surveyed the key properties of scrambling in Korean. In-depth studies on Korean scrambling contribute to the understanding of the typology of movement and displacement in general. Korean scrambling cannot be directly mapped into the dichotomy of A/A'-movement. The current observation naturally leads to the pursuit of a general theory, which covers obligatory movement as well as optional displacement with mixed semantic effects. The contrasts between Korean and other scrambling languages engender an important research question on how to account for cross-linguistic differences and similarities among languages with free word order. Investigations on restrictions in scrambling tell us how the syntax interacts with various components of the grammar at the interfaces.


Aarts, B. (1992). Small clauses in English: The nonverbal types. Berlin: Mouton de Gruyter.Find this resource:

Abe, J. (1993). Binding conditions and scrambling without A/A' distinction (Unpublished doctoral dissertation). University of Connecticut, Storrs.Find this resource:

Ahn, S.-H. (1990). Korean quantification and universal grammar (Unpublished doctoral dissertation). University of Connecticut, Storrs.Find this resource:

Aoun, J., & Li, Y.-H. A. (1989). Scope and constituency. Linguistic Inquiry, 20, 141–172.Find this resource:

Bailyn, J. F. (2001). On scrambling: A reply to Bošković and Takahashi. Linguistic Inquiry, 32, 635–658.Find this resource:

Barss, A. (1986). Chains and anaphoric dependence: On reconstruction and its implications (Unpublished doctoral dissertation). Cambridge, MA: MIT.Find this resource:

Beck, S., & Kim, S. -S. (1997). On wh-and operator scope in Korean. Journal of East Asian Linguistics, 6, 339–384.Find this resource:

Belletti, A., & Rizzi, L. (1988). Psych-verbs and θ‎-theory. Natural Language & Linguistic Theory, 6, 291–352.Find this resource:

Bošković, Ž. (2004). Topicalization, focalization, lexical insertion, and scrambling. Linguistic Inquiry, 35, 613–638.Find this resource:

Bošković, Ž., & Takahashi, D. (1998). Scrambling and last resort. Linguistic Inquiry, 29, 347–366.Find this resource:

Cho, J.-H. (1994a). Scrambling in Korean: Crossover, reconstruction and binding theory (Unpublished doctoral dissertation). University of Connecticut, Storrs.Find this resource:

Cho, J.-H. (1994b). On scrambling: reconstruction, crossover, and anaphor binding. In Y.-K. Kim-Renaud (Ed.), Theoretical issues in Korean linguistics (pp. 255–274). Stanford, CA: CSLI.Find this resource:

Cho, J.-H. (1996). Reconstruction effects and intermediate A-position. Studies in Generative Grammar, 6, 257–279.Find this resource:

Cho, J.-H., & Kim, O.-H. (2000). Scrambling without syntactic movement. Studies in Generative Grammar, 10, 151–179.Find this resource:

Choe, H.-S. (1985). Remarks on configurationality parameters. In S. Kuno, J. Whitman, I.-H. Lee, & Y.-S. Kang (Eds.), Harvard studies in Korean linguistics I (pp. 14–29). Department of Linguistics, Harvard University, Cambridge, MA.Find this resource:

Choi, K. (2014). Phikyelsok-pyenhang taymyengsaloseuy ‘caki’ [caki as a bound-variable pronoun]. Studies in Generative Grammar, 24(2), 325–364.Find this resource:

Choi, H.-W. (1999). Optimizing structure in context: Scrambling and information structure. Stanford, CA: CSLI Publications.Find this resource:

Choi, Y.-S. (2004a). Reconstruction, weak crossover, and binding in Korean scrambling: Some theoretical implications. The Linguistic Association of Korea Journal, 12(2), 185–206.Find this resource:

Choi, Y.-S. (2004b). The structure of selo and its implication for binding theory. Language Research, 40(3), 681–694.Find this resource:

Choi, Y.-S. (2007). Intervention effect in Korean wh-questions: Indefinite and beyond. Lingua, 117, 2055–2076.Find this resource:

Chomsky, N. (1965). Aspects of the theory of syntax. Cambridge, MA: MIT Press.Find this resource:

Chomsky, N. (1981). Lectures on government and binding. Dordrecht: Foris.Find this resource:

Chomsky, N. (1982). Some concepts and consequences of the theory of government and binding. Cambridge, MA: MIT Press.Find this resource:

Chomsky, N. (1986). Barriers. Cambridge, MA: MIT Press.Find this resource:

Chomsky, N. (1995). The minimalist program. Cambridge, MA: MIT Press.Find this resource:

Chomsky, N. (2000). Minimalist inquiries: The framework. In R. Martin, D. Michaels, & J. Uriagereka (Eds.), Step by step: Essays on Minimalist syntax in honor of Howard Lasnik (pp. 89–155). Cambridge, MA: MIT Press.Find this resource:

Chomsky, N. (2001). Derivation by phase. In M. Kenstowicz (Ed.), Ken Hale: A life in language (pp. 1–52). Cambridge, MA: MIT Press.Find this resource:

Chomsky, N. (2008). On phases. In R. Freidin, C. P. Otero, & M. L. Zubizarreta (Eds.), Foundational issues in linguistic theory: Essays in honor of Jean-Roger Vergnaud (pp. 133–166). Cambridge, MA: MIT Press.Find this resource:

Chomsky, N. (2013). Problems of projection. Lingua, 130, 33–49.Find this resource:

Chomsky, N., & Lasnik, H. (1977). Filters and control. Linguistic Inquiry, 8, 425–504.Find this resource:

Chung, D. (1996). On the representation and licensing of Q and Q-dependents (Unpublished doctoral dissertation). University of Southern California, Los Angeles.Find this resource:

Chung, D. (2008). Agree but not necessarily at the same time. Studies in Generative Grammar, 18, 509–524.Find this resource:

Chung, D., & Park, H.-K. (1998). On selo. In D. J. Silva (Ed.), Japanese/Korean linguistics (8th ed., pp. 417–430). Stanford, CA: CSLI Publications.Find this resource:

Corver, N., & van Riemsdijk, H. (Eds.). (1994). Studies on scrambling: Movement and non-movement approaches to free word-order phenomena. Berlin: Walter de Gruyter.Find this resource:

den Dikken, M. (2006). Relators and linkers: The syntax of predication, predicate inversion, and copulas. Cambridge, MA: MIT Press.Find this resource:

den Dikken, M. (2007). Phase extension: Contours of a theory of the role of head movement in phrasal extraction. Theoretical Linguistics, 33, 1–41.Find this resource:

Epstein, S. D., Groat, E. M., Kawashima, R., & Kitahara, H. (1998). A derivational approach to syntactic relations. Oxford: Oxford University Press.Find this resource:

Fanselow, G. (2001). Features, θ‎-roles, and free constituent order. Linguistic Inquiry, 32, 405–437.Find this resource:

Fiengo, R. (1977). On trace theory. Linguistic Inquiry, 8, 35–62.Find this resource:

Fox, D. (1999). Reconstruction, binding theory, and the interpretation of chains. Linguistic Inquiry, 30, 157–196.Find this resource:

Fox, D. (2000). Economy and semantic interpretation. Cambridge, MA: MIT Press.Find this resource:

Fox, D., & Pesetsky, D. (2005). Cyclic linearization of syntactic structure. Theoretical Linguistics, 31, 1–45.Find this resource:

Fukui, N. (1993). Parameters and optionality. Linguistic Inquiry, 24, 399–420.Find this resource:

Grewendorf, G., & Sabel, J. (1999). Scrambling in German and Japanese: Adjunction versus multiple specifiers. Natural Language and Linguistic Theory, 17, 1–65.Find this resource:

Hagstrom, P. (1998). Decomposing questions (Unpublished doctoral dissertation). MIT, Cambridge, MA.Find this resource:

Hale, K. (1982). Preliminary remarks on configurationality. In Proceedings of NELS (Vol. 12, pp. 86–96). GLSA, Department of Linguistics, University of Massachusetts, Amherst.Find this resource:

Han, C.-h., & Storoshenko, D. R. (2012). Semantic binding of long-distance anaphor caki in Korean. Language, 88(4), 764–790.Find this resource:

Hoji, H. (1985). Logical form constraints and configurational structures in Japanese (Unpublished doctoral dissertation). University of Washington, Seattle.Find this resource:

Huang, C.-T. J. (1982). Logical relations in Chinese and the theory of grammar (Unpublished doctoral dissertation). MIT, Cambridge, MA.Find this resource:

Johnston, J. C., & Park, I. (2001). Some problems with a lowering account of scrambling. Linguistic Inquiry, 32, 727–732.Find this resource:

Jung, Y.-J. (2002). Scrambling, edge effects and A/A'-distinction. The Linguistic Association of Korea Journal, 10, 41–64.Find this resource:

Jung, Y.-J. (2015). Subjacency revisited: Is it real or a reflex of performance biases? Korean Journal of Linguistics, 40, 511–542.Find this resource:

Kang, B.-M. (1998). Three kinds of Korean reflexives: A corpus linguistic investigation on grammar and usage. In J. Guo, K. T. Lua, & J. Xu (Eds.), Proceedings of the 12th Pacific Asia Conference on Language, Information, and Computation (pp. 10–19). Singapore: PACLIC.Find this resource:

Kang, J.-G., & Müller, G. (1996). Reconstruction vs. copying: The case of wh-scope. In N. Akatsuka, S. Iwasaki, & S. Strauss (Eds.), Japanese/Korean linguistics (5th ed., pp. 269–285). Stanford, CA: CSLI Publications.Find this resource:

Karimi, S. (1999). Is scrambling as strange as we think it is? In K. Arregi, B. Bruening, C. Krause, & V. Lin (Eds.), Papers on morphology and syntax: Cycle one (pp. 159–190). MIT Working Papers in Linguistics 33. Cambridge, MA: MITWPL.Find this resource:

Karimi, S. (2005). A minimalist approach to scrambling: Evidence from Persian. Berlin: Mouton de Gruyter.Find this resource:

Karimi, S. (Ed.). (2003). Word order and scrambling. Malden, MA: Blackwell.Find this resource:

Kempson, R., & Kiaer, J. (2010). Multiple long-distance scrambling: Syntax as reflection of processing. Journal of Linguistics, 46, 127–192.Find this resource:

Kim, D.-B. (1993). The specificity/non-specificity distinction and scrambling theory. Seoul: Taehaksa.Find this resource:

Kim, D.-B. (1995). Non-specificity and scrambling. Studies in Generative Grammar, 5, 199–238.Find this resource:

Kim, J.-H., & Yoon, J. H. (2009). Long-distance bound local anaphors in Korean: An empirical study of the Korean anaphor caki-casin. Lingua, 119, 733–755.Find this resource:

Kim, J.-H., Montrul, S., & Yoon, J. H. (2009). Binding interpretation of anaphors by Korean heritage speakers. Language Acquisition, 16, 3–35.Find this resource:

Kim, L. (2015). Asymmetric ditransitive constructions: Evidence from Korean. Lingua, 165, 28–69.Find this resource:

Kim, R.-H.-Y. (2003). Scrambling and minimalist program. Studies in Generative Grammar, 13, 3–19.Find this resource:

Kim, S.-S. (2002). Intervention effects are focus effects. In N. Akatsuka & S. Strauss (Eds.), Japanese/Korean linguistics (10th ed., pp. 615–628). Stanford, CA: CSLI Publications.Find this resource:

Kim, S. (1991). Chain scope and quantification structure (Unpublished doctoral dissertation). Brandeis University, Waltham, MA.Find this resource:

Kim, S. (1999). Sloppy/strict identity, empty objects, and NP ellipsis. Journal of East Asian Linguistics, 8, 255–284.Find this resource:

Kim, Y.-H. (1998). Overt Case and covert Case in Korean. Studies in Generative Grammar, 8, 177–237.Find this resource:

Kim, Y.-H. (1999). Hankwuke kyekkwa eswunuy choysocwuuy mwunpep. Seoul: Hankwukmwunhwasa.Find this resource:

Kim, Y.-S. (1997). The LF analysis of scrambling. Studies in Modern Grammatical Theories, 11, 79–90.Find this resource:

Kim, Y.-J., & Larson, R. (1989). Scope interpretation and the syntax of psych-verbs. Linguistic Inquiry, 20(4), 681–688.Find this resource:

Kitahara, H. (2002). Scrambling, case, and interpretability. In S. D. Epstein & T. D. Seely (Eds.), Derivation and explanation in the minimalist program (pp. 167–183). Malden, MA: Blackwell.Find this resource:

Ko, H. (2005a). Syntactic edges and linearization (Unpublished doctoral dissertation). MIT, Cambridge, MA.Find this resource:

Ko, H. (2005b). Syntax of why-in-situ: Merge into [Spec,CP] in the overt syntax. Natural Language and Linguistic Theory, 23, 867–916.Find this resource:

Ko, H. (2007). Asymmetries in scrambling and cyclic linearization. Linguistic Inquiry, 38, 49–83.Find this resource:

Ko, H. (2011). Predication and edge effects. Natural Language and Linguistic Theory, 29, 725–778.Find this resource:

Ko, H. (2014a). Edges in syntax: Scrambling and cyclic linearization. Oxford: Oxford University Press.Find this resource:

Ko, H. (2014b). Remarks on right dislocation construction in Korean: Challenges to biclausal analyses. Language Research, 50, 275–310.Find this resource:

Ko, H. (2015). On the typology of small clauses: Null subject and mode of merge in resultatives. Studies in Generative Grammar, 25, 347–375.Find this resource:

Ko, H. (2016). Gapless right-dislocation: The role of overt correlates. Language Research, 52, 1–30.Find this resource:

Kwon, S.-N., & Zribi-Hertz, A. (2008). Differential function marking, case, and information structure: Evidence from Korean. Language, 84(2), 258–299.Find this resource:

Kuno, S. (1973). The structure of the Japanese language. Cambridge, MA: MIT Press.Find this resource:

Kuno, S. (1980). A further note on Tonoike’s intra-subjectivization hypothesis. In Y. Otsu & A. Farmer (Eds.), Theoretical issues in Japanese linguistics (pp. 171–184). MIT Working Papers in Linguistics 2. Cambridge, MA: MITWPL.Find this resource:

Kuroda, S.-Y. (1988). Whether we agree or not: A comparative syntax of English and Japanese. Lingvisticae Investigationes, 12, 1–47.Find this resource:

Kwon, J. (2010). Scrambling of Korean embedded clauses. Studies in Generative Grammar, 20, 229–249.Find this resource:

Lasnik, H. (1999). Chains of arguments. In S. D. Epstein & N. Hornstein (Eds.), Working minimalism (pp. 189–215). Cambridge, MA: MIT Press.Find this resource:

Lasnik, H., & Stowell, T. (1991). Weakest crossover. Linguistic Inquiry, 22, 687–720.Find this resource:

Lebeaux, D. (1988). Language acquisition and the form of the grammar (Unpublished doctoral dissertation). University of Massachusetts, Amherst.Find this resource:

Lebeaux, D. (2009). Where does binding theory apply? Cambridge, MA: MIT Press.Find this resource:

Lee, H. (2015). Case particle ellipsis. In L. Brown & J. Yeon (Eds.), Handbook of Korean linguistics (pp. 196–211). Hoboken, NJ: John Wiley & Sons.Find this resource:

Lee, H. (2006). A study on scrambling in Korean: The minimalist approach. Studies in Generative Grammar, 16, 425–460.Find this resource:

Lee, H. (2001a). Interpreting selo as a variable: The syntax and semantics interface. Language Research, 37, 293–309.Find this resource:

Lee, H. (2001b). A case of anaphor-antecedent reversal and binding theory. Eoneohak, 29, 133–152.Find this resource:

Lee, H. (2006). A comparative study of reciprocals in English and Korean: Their distribution and interpretation. Studies in Generative Grammar, 16(1), 129–152.Find this resource:

Lee, J.-E. (2004). Ditransitive structures and (anti-)locality (Unpublished doctoral dissertation). Harvard University, Cambridge, MA.Find this resource:

Lee, K.-H. (2003). Semantic effects in scrambling. In G. K. Iverson & S.-C. An (Eds.), Explorations in Korean language and linguistics (pp. 377–390). Seoul: Hankook.Find this resource:

Lee, W., & Cho, S. (2003a). Argument scrambling and object shift. Studies in Generative Grammar, 13, 39–59.Find this resource:

Lee, W., & Cho, S. (2003b). Is scrambling EPP-driven? Studies in Generative Grammar, 13, 331–341.Find this resource:

Lee, Y.-S. (1993). Scrambling as case-driven obligatory movement (Unpublished doctoral dissertation). University of Pennsylvania.Find this resource:

Lee, Y.-S. (1994). Scrambling as case-driven movement. Studies in Generative Grammar, 4, 495–533.Find this resource:

Madigan, S. (2015). Anaphora and binding. In L. Brown & J. Yeon (Eds.), Handbook of Korean linguistics (pp. 137–154). Hoboken, NJ: John Wiley & Sons.Find this resource:

Mahajan, A. K. (1990). The A/A'-distinction and movement theory (Unpublished doctoral dissertation). MIT, Cambridge, MA.Find this resource:

Martin, S. E. (1992). A reference grammar of Korean: A complete guide to the grammar and history of the Korean language. Rutland, VT: Charles E. Tuttle.Find this resource:

May, R. (1977). The grammar of quantification (Unpublished doctoral dissertation). MIT, Cambridge, MA.Find this resource:

May, R. (1985). Logical form: Its structure and derivation. Cambridge, MA: MIT Press.Find this resource:

Miyagawa, S. (1989). Structure and case marking in Japanese (Vol. 22). Cambridge, MA: Academic Press.Find this resource:

Miyagawa, S. (1997). Against optional scrambling. Linguistic Inquiry, 28, 1–25.Find this resource:

Miyagawa, S. (2001). The EPP, scrambling, and wh-in-situ. In M. Kenstowicz (Ed.), Ken Hale: A life in language (pp. 293–338). Cambridge, MA: MIT Press.Find this resource:

Miyagawa, S. (2006). On the ‘undoing’ nature of scrambling: A response to Bošković. Linguistic Inquiry, 37, 607–624.Find this resource:

Miyagawa, S. (2010). Why Agree? Why Move? Unifying agreement–based and discourse-configurational languages. Cambridge, MA: MIT Press.Find this resource:

Miyagawa, S. (2017). Agreement beyond phi. Cambridge, MA: MIT Press.Find this resource:

Neeleman, A., & Reinhart, T. (1998). Scrambling and the PF interface. In M. Butt & W. Geuder (Eds.), The projection of arguments: Lexical and compositional factors (pp. 309–353). Stanford, CA: CSLI Publications.Find this resource:

Nemoto, N. (1993). Chains and case positions: A study from scrambling in Japanese (Unpublished doctoral dissertation). University of Connecticut, Storrs.Find this resource:

Nishigauchi, T. (2002). Scrambling and reconstruction at LF. Gengo Kenkyu, 121, 49–105.Find this resource:

Postal, P. (1971). Cross-over phenomena. New York: Holt, Rinehart, and Winston.Find this resource:

Reinhart, T. (1981). Definite NP anaphora and c-command domains. Linguistic Inquiry, 12, 605–635.Find this resource:

Richards, N. (2001). Movement in language: Interactions and architecture. Oxford: Oxford University Press.Find this resource:

Richards, N. (2014). A-bar movement. In A. Carnie, Y. Sato, & D. Siddiqi (Eds.), The Routledge handbook of syntax (pp. 167–191). New York: Routledge.Find this resource:

Rizzi, L. (1990). Relativized minimality. Cambridge, MA: MIT Press.Find this resource:

Ross, J. R. (1967). Constraints on variables in syntax (Unpublished doctoral dissertation). MIT, Cambridge, MA.Find this resource:

Sabel, J. (2001). Wh-questions in Japanese: Scrambling, reconstruction, and wh-movement. Linguistic Analysis, 31, 1–41.Find this resource:

Sabel, J. (2005). String-vacuous scrambling and the effect on output condition, In J. Sabel & M. Saito (Eds.), The free word order phenomenon: Its syntactic sources and diversity (pp. 281–334). Berlin: Mouton de Gruyter.Find this resource:

Sabel, J., & Saito, M. (Eds.). (2005). The free word order phenomenon: Its syntactic sources and diversity. Berlin: Mouton de Gruyter.Find this resource:

Saito, M. (1985). Some asymmetries in Japanese and their theoretical implications (Unpublished doctoral dissertation). MIT, Cambridge, MA.Find this resource:

Saito, M. (1989). Scrambling as semantically vacuous A'-movement. In M. R. Baltin & A. S. Kroch (Eds.), Alternative conceptions of phrase structure (pp. 182–200). Chicago: University of Chicago Press.Find this resource:

Saito, M. (1992). Long-distance scrambling in Japanese. Journal of East Asian Linguistics, 1, 69–118.Find this resource:

Saito, M. (2003). A derivational approach to the interpretation of scrambling chains. Lingua, 113, 481–518.Find this resource:

Saito, M., & Hoji, H. (1983). Weak crossover and move α‎ in Japanese. Natural Language and Linguistic Theory, 1, 245–259.Find this resource:

Saito, M., & Fukui, N. (1998). Order in phrase structure and movement. Linguistic Inquiry, 29, 439–474.Find this resource:

Sato, Y., & Goto, N. (2014). Scrambling. In A. Carnie, Y. Sato, & D. Siddiqi (Eds.), The Routledge handbook of syntax (pp. 264–282). New York: Routledge.Find this resource:

Sauerland, U., & Elbourne, P. (2002). Total reconstruction, PF movement, and derivational order. Linguistic Inquiry, 33, 283–319.Find this resource:

Sells, P. (2015). Negation and negative polarity items. In L. Brown & J. Yeon (Eds.), Handbook of Korean linguistics (pp. 212–231). Hoboken, NJ: John Wiley & Sons.Find this resource:

Sohn, K.-W. (1995). Negative polarity items, scope, and economy (Unpublished doctoral dissertation). University of Connecticut, Storrs.Find this resource:

Son, G. (2001). Scrambling, reconstruction, and the checking principle (Unpublished doctoral dissertation). University of Wisconsin-Madison.Find this resource:

Son, G. (2003). Traces of wh-scrambling. Studies in Generative Grammar, 13, 99–118.Find this resource:

Song, J. J. (2009). Word order patterns and principles: An overview. Language and Linguistics Compass, 3(5), 1328–1341.Find this resource:

Song, J. J. (2012). Word order. Cambridge, U.K.: Cambridge University Press.Find this resource:

Suh, J. (1990). Scope phenomena and aspects of Korean syntax (Unpublished doctoral dissertation). University of Southern California, Los Angeles.Find this resource:

Tada, H. (1993). A/A-bar partition in derivation (Unpublished doctoral dissertation). MIT, Cambridge, MA.Find this resource:

Takahashi, D. (1993). Movement of wh-phrases in Japanese. Natural Language and Linguistic Theory, 11, 655–678.Find this resource:

Takahashi, S., & Hulsey, S. (2009). Wholesale merger: Beyond the A/A'-distinction. Linguistic Inquiry, 40, 387–426.Find this resource:

Takano, Y. (1998). Object shift and scrambling. Natural Language and Linguistic Theory, 16, 817–889.Find this resource:

Tomioka, S. (2007). Pragmatics of LF intervention effects: Japanese and Korean wh-interrogatives. Journal of Pragmatics, 39, 1570–1590.Find this resource:

Tsai, W.-T. D. (1994). On economizing the theory of A-bar dependencies (Unpublished doctoral dissertation). MIT, Cambridge, MA.Find this resource:

Uriagereka, J. (1999). Multiple spell-out. In S. Epstein & N. Hornstein (Eds.), Working minimalism (pp. 251–282). Cambridge, MA: MIT Press.Find this resource:

Watanabe, A. (1992). Wh-in-situ, subjacency, and chain formation. MIT Occasional Papers in Linguistics 2. Cambridge, MA: MITWPL.Find this resource:

Webelhuth, G. (1989). Syntactic saturation phenomena and the modern Germanic languages (Unpublished doctoral dissertation). University of Massachusetts, Amherst.Find this resource:

Whitman, J. (1987). Configurationality parameters. In T. Imai & M. Saito (Eds.), Issues in Japanese linguistics (pp. 351–374). Dordrecht: Foris.Find this resource:

Yang, D.-W. (1983). The extended binding theory of anaphors. Language Research, 19(2), 169–192.Find this resource:

Yang, D.-W. (1986). Hankwuke-uy tayyongsalon. Kwukehak, 15, 41–162.Find this resource:

Yang, D.-W., & Kim, S.-H. (2005). Scrambling and interpretive complex. The Linguistic Association of Korea Journal, 13(2), 169–193.Find this resource:

Yoon, J. H. (2004). Non-nominative (major) subjects and case stacking in Korean. In P. Bhaskararao & K. V. Subbarao (Eds.), Non-nominative subjects (Vol. 2, pp. 265–314). Typological Studies in Language 61. Amsterdam: John Benjamins.Find this resource:

Yoon, J. H. (2007). Raising of major arguments in Korean and Japanese. Natural Language and Linguistic Theory, 25, 615–653.Find this resource:

Yoon, J. H. (2015). Double nominative and double accusative constructions. In L. Brown & J. Yeon (Eds.), Handbook of Korean linguistics (pp. 79–97). Hoboken, NJ: John Wiley & Sons.Find this resource:

Yoon, J.-Me. (2013). Undoing and wh-Island effects of scrambling in Korean. Studies in Generative Grammar, 23, 41–63.Find this resource:

Zubizarreta, M. L. (1998). Prosody, focus and word order. Cambridge, MA: MIT Press.Find this resource:


(1.) Word order variation in Korean has been attributed to various syntactic operations, such as leftward scrambling, topicalization, left-dislocation, and right-dislocation (also called rightward scrambling/afterthought). Among these, this article focuses on the syntax and semantics of leftward scrambling, which is one of the most discussed and prominent topics in Korean linguistics. The literature covered in this article is largely confined to previous works developed under the Generative Grammar framework (in particular, the Government and Binding Theory, the Minimalist Program, and its predecessors that cohere with the Principles and Parameters approach to language, e.g., Chomsky, 1965, 1981, 1982, 1986, 1995, 2001, 2008, 2013). See Song (2009, 2012) for a critical survey of different approaches to word order, based on linguistic typology, generative grammar, optimality theory, and performance-based theories. This article employs the Yale Romanization (Martin, 1992) to transliterate Korean examples. The back vowel [u], however, is consistently transliterated as ‘wu’ regardless of its phonetic context.

(2.) The term reconstruction is employed in this article to refer to interpretive effects such that a phrase is interpreted in non-surface position (most likely, the base position) after overt movement. This article abstracts away from the theoretical discussion on whether reconstruction implies literal lowering at LF (May, 1977, 1985) or interpretation of a (part of) lower copy at LF (Chomsky, 1995).

(3.) Most examples in Section 2.1 are drawn from the literature on scrambling in Korean (with minor modifications in notations, styles, and transliteration). To clarify, some of the Korean examples are modeled after data studied in other languages; see, in particular, Mahajan (1990, Hindi), Saito (1992, Japanese), and Tada (1993, Japanese) for tests concerning binding and WCO effects. Given that the notion of A- and A-bar distinction can be applicable to language in general, it is not surprising that researchers tested A/A'-effects in different languages with similar data structures. If relevant, the article refers the reader to the literature on other languages in the text. This, however, does not mean that all the scrambling languages behave in the same way on the test at issue. For data in other languages, refer to primary sources cited for each language.

(4.) Y. Choi (2004a, pp. 189–190) reports that Korean data similar to (8) are grammatical, however. The reader may note that the data in (8) (also (12)) may be explained by assuming that the subject in fact c-commands the scrambled R-expression adjoined in the same TP (under c-command in the sense of Reinhart, 1981). Some previous studies have taken such an approach (see Bošković & Takahashi, 1998, p. 361 for Japanese, and Son, 2001, Chapter 2 for Korean scrambling and binding under an IP-adjunction structure). It is unclear, however, whether this approach would explain variations in Condition C (reconstruction) effects in languages. Bailyn (2001, p. 643) reports that Condition C (reconstruction) effects hold only when a certain semantic condition is met for the scrambled NP in Russian. Citing Nishigauchi (2002), Miyagawa (2006) also argues that reconstruction effects due to Condition C are not observed in Japanese when an R-expression is embedded under a scrambled adjunct phrase. These variations in reconstruction are not expected under the adjunction-based approach to Condition C. It remains open whether Condition C effects in Korean are subject to similar variations reported for other languages and whether the data in Korean support the adjunction-based approach over the standard LF-reconstruction approach.

(5.) Based on Hindi, Mahajan (1990) argued that clause-internal scrambling may target the A or A'-position, while clause-external scrambling must target the A'-position. Saito (1992) argues that Japanese scrambling shows a property of non-operator movement at S-structure (following Webelhuth’s (1989) approach to German) but must be reanalyzed as A or A'-movement at LF (adapting Mahajan’s (1990) analysis of Hindi). Under Saito (1992), clause-internal scrambling can be reanalyzed as either A or A'-movement (depending on verb raising), whereas clause-external scrambling must be reanalyzed as A'-movement at LF. See Sato and Goto (2014) for an overview on Japanese scrambling.

(6.) The reader may note that wh-phrases and NPIs (e.g., (13), (14), (17)) may be licensed at any point of derivation without assuming reconstruction at LF (see Sohn, 1995 for a related discussion on NPI licensing and feature checking in Korean). It is unclear, however, how this alternative would explain wh-licensing in (18). Once a wh-feature is checked in the embedded clause, it will not be checked again in the matrix clause. Thus, the ambiguity claimed for (18), if it exists, would not be explained in any obvious way under the derivational checking account. Moreover, a wide range of approaches have been proposed for wh-licensing in in-situ languages, and thus a careful study should be done to evaluate such a derivational approach to wh-licensing (cf. Huang, 1982; Watanabe, 1992; Takahashi, 1993; Tsai, 1994; Chung, 1996, 2008; Hagstrom, 1998; Jung, 2015, among many others).

(7.) More precisely, Sohn (1995, pp. 151, 199–200) claims that if manhun ‘many’ undergoes scrambling to sentential initial position (over an NPI and negation), it must be interpreted as a specific group whose cardinality is many. This generalization holds in both clause-external and clause-internal scrambling. Sohn interprets the contrast between (19a) and (19b) to indicate that not all instances of long-distance scrambling are undone, contra Tada (1993) (see also Bailyn, 2001, p. 642, and Miyagawa, 2006 for anti-reconstruction effects in long-distance scrambling). It should be noted, however, that Sohn acknowledges reconstruction effects of long-distance scrambling in other cases. Sohn (1995, p. 188) reports that when negation is not involved, as in (i), manhun ‘many’ must be reconstructed back after long-distance scrambling and can take scope only in the embedded clause (as argued for Japanese in Tada, 1993). It remains an open question for future research when and why a scrambled quantifier must be interpreted as a specific indefinite, as in (19a), or as a quantificational element with LF-reconstruction effects, as in (i).


Scrambling in Korean Syntax

(8.) See also Y. Lee (1994) for the lack of reconstruction effects in pronominal (variable) binding in VP-internal scrambling. A cautionary note is in order in the interpretation of (23), however. In contrast to caki in (23), some speakers find that a reciprocal DO can be bound by an IO, as in (i). Given this, one might argue that (23) is ungrammatical because caki is strongly subject-oriented in Korean, and that the DO in (i) may undergo reconstruction below the IO, contra Cho (1994a, 1994b) (see Miyagawa, 1997 for the claim on Japanese that DO may undergo A'-scrambling over IO and subsequent reconstruction below IO when focus is involved). See also note 11 for further discussion on anaphors in Korean.


Scrambling in Korean Syntax

(9.) Y. Lee (1993) claims that adjunct scrambling and clause-external scrambling are Case-driven as well. To explain counterexamples such as (6) (in conjunction with (8)), Y. Lee (1994, p. 528) assumes that Condition C applies at SS and cannot be “undone” at LF, which amounts to saying that Condition C must be satisfied in the scrambled position as well as in the reconstructed/original position.

(10.) The idea that scrambling can be a type of non-operator A'-movement dates back to earlier influential works by Webelhuth (1989) and Saito (1992) (see note 5). Cho (1994a, 1994b) adapts this idea to clause-internal and external scrambling in Korean that occurs across a subject. Departing from Saito (1992), who assumes LF-reanalysis of scrambling, however, Cho does not postulate LF-reanalysis but assumes that non-operator scrambling must be understood as non-operator A'-movement throughout the derivation.

(11.) It has been reported that some reflexives in Korean (e.g., caki, casin) can be used as a long-distance anaphor quite freely, while other reflexives (e.g., caki-casin, ku casin) strongly prefer a local (clause-bound) reading (see Kang, 1998; Kim, Montrul, & Yoon, 2009; Madigan 2015; cf. Kim & Yoon, 2009 for caki-casin as a logophor). It is rather controversial how these reflexives should be analyzed. Various suggestions have been made for reflexives in Korean (e.g., local anaphor, long-distance anaphor, pronoun, logophor/exempt anaphor, bound variable). See, for instance, Han and Storoshenko (2012) and K. Choi (2014) for reviews and controversies on caki. Kim and Yoon (2009, p. 754) concludes that there are no anaphors that are used exclusively for core binding (in the sense of Chomsky, 1986), with others being reserved for discourse-bound exempt binding. Instead of reflexives, some researchers used selo ‘each other’ in testing reconstruction effects. The status of selo, however, is still unclear, and diverse suggestions have been made for selo as well (e.g., local anaphor, bound variable/resumptive pronoun, coreferential pronominal, and collective adverb). See Yang 1986; Ahn 1990; Chung and Park 1998; H. Lee 2001a, 2001b, 2006; and Y. Choi, 2004b for discussions on selo. Overall, current debates on anaphors lead us to the point that anaphor binding in Korean cannot be taken as straightforward evidence for (anti-)reconstruction effects which have been observed with core binding in other languages. Only after the nature of anaphor binding in Korean is properly understood can its consequence for scrambling be properly evaluated.

(12.) Alternatively, one may rule out (35c) by extending the anti-ambiguity constraint employed for (33a) (cf. Kuno, 1980; Y. Lee, 1993). The linear order in (35a) may be potentially ambiguous in syntax, but the canonical parse is favored over the one with string-vacuous scrambling, and thus the derivation in (35c) becomes unavailable.

(13.) Judgments on (39) and (40) are controversial, however. Some speakers find that (39) and (40) are acceptable when the Caseless NP is interpreted as a left-dislocated topic or hanging topic, and some speakers also find that (40) becomes acceptable when the subject is marked with nominative Case instead of the Topic marker.

(14.) Under cyclic syntax (e.g., Uriagereka, 1999; Chomsky, 2001; Fox & Pesetsky, 2005), all the syntactic structures must be cyclically linearized at the interface after Spell-out. The consequences of cyclic Spell-out are understood differently from theory to theory, however. Ko (2014a) adopts Fox and Pesetsky’s (2005) CL model over Chomsky’s (2001) phase model and derives peculiar properties of syntactic edges from various interactions among scrambling, predicational structures, and CL. Refer to Ko (2014a) for further discussion on scrambling and cyclic syntax. See also Ko (2014a) and references therein for other important issues such as the cross-linguistic typology and structure of numeral quantifiers and the role of A/A'-scrambling in floating quantifier constructions.