1 Introduction and Background

It has long been assumed that argument wh-in-situ in Chinese is not sensitive to islands, unlike adjunct wh-in-situ (Huang 1982a,b, Tsai 1994a,b, 1999, Cheng 2009).1 On the basis of a formal acceptability-rating experiment (Sprouse 2007, Sprouse, Wagers, and Phillips 2012, Sprouse and Hornstein 2013), this squib shows that, contrary to this long-standing generalization, both argument and adjunct wh-in-situ are sensitive to complex NP islands (Ross 1967). We argue that our finding is most compatible with the theory of wh-in-situ whereby both argument and adjunct wh-phrases undergo a similar movement process, namely, phrasal movement.

Studies of wh-in-situ in Chinese have identified the following generalization, known as the argument-adjunct asymmetry: argument wh-in-situ (e.g., shenme ‘what’, shei ‘who’) is not sensitive to islands, but adjunct wh-in-situ (e.g., weishenme ‘why’, zenmeyang ‘how’) is severely constrained by islands (Huang 1982a,b, Xu 1990, Lasnik and Saito 1992, Tsai 1994a,b, 1999). For example, in (1a) and (1b), respectively, a wh-DP (shenme ‘what’) and a wh-adjunct (weishenme ‘why’) are embedded within a relative clause (the Complex NP Constraint; Ross 1967).2 It has been reported that (1a) is more acceptable than (1b) (examples from Tsai 1999:42–43).3

(1)

  • Argument

    Akiu kan-bu-qi [NP[CP ei zuo shenme de] reni]?

    Akiu look-not-up [NP[CP ei do what REL] personi]

    ‘What is the thing/job x such that Akiu despises [people [who do x]]?’

  • Adjunct

    *Akiu xihuan [NP[CP Luxun weishenme xie ei de]

    Akiu like [NP[CP Luxun why write eiREL] shui]?

    booki]

    ‘What is the reason x that Akiu likes [books [that Luxun wrote for x]]?’

Part of the syntactic and semantic research into Chinese wh-in-situ has thus been devoted to explaining the argument-adjunct asymmetry shown in (1). Let us look at some of the dominant approaches to Chinese wh-in-situ.

One approach attributes the contrast in (1) to the Empty Category Principle (ECP) (Chomsky 1981, 1986, Huang 1982b, Lasnik and Saito 1984, 1992, Cheng 2009). It has been suggested that the Subjacency Condition (Chomsky 1977, 1981, 1986) is an S-Structure condition (or a constraint on overt phrasal movement), whereas the ECP is an LF condition (or a constraint on covert movement) (Lasnik and Saito 1984, 1992). Since Chinese wh-in-situ does not undergo overt wh-movement, the Subjacency Condition should not be in effect. However, Chinese wh-in-situ should be sensitive to the ECP if it undergoes movement in LF. Under the ECP-based explanation, the trace left by an adjunct wh-element is not properly governed (specifically, an intermediate trace is not antecedent-governed; see Chomsky 1981, 1986, Lasnik and Saito 1992) and thus violates the ECP (Huang 1982a,b, Lasnik and Saito 1984, 1992). Following this argument, we would expect adjunct but not argument wh-in-situ to be sensitive to islands.4

More recent studies have suggested that argument wh-in-situ does not undergo movement and is therefore immune from island effects, whereas adjunct wh-in-situ undergoes covert movement and is thus sensitive to islands (Aoun and Li 1993, Chomsky 1995, Tsai 1999, Stepanov and Tsai 2008). Tsai (1999), for example, argues that wh-DPs are subject to unselective binding whereas wh-PPs are not. Thus, wh-DPs can create operator-variable pairs by means of unselective binding without resorting to wh-movement. Wh-PPs, on the other hand, cannot create operator-variable pairs by means of unselective binding and must undergo movement to do so. Stepanov and Tsai (2008) further propose a distinction between two interpretations of the Chinese wh-element weishenme (‘purpose why’/‘for what’ and ‘reason why’/ ‘why’) and claim that unselective binding is available for ‘for what’ but not for ‘why’. Therefore, a certain interpretation of weishenme prompts movement of the wh-element, to which island constraints should apply.

Other studies have explained the island sensitivity of wh-in-situ by introducing movements of a question feature, formulated either as a “wh-feature” (Pesetsky 2000), a possibly null question particle that undergoes “Q-movement” in overt syntax (Watanabe 1992, Hagstrom 1998), or a “Qu-operator” base-generated in a “Qu-projection” and bound by wh-in-situ (Aoun and Li 1993). The movement of such question particles (or features) to Spec,CP is constrained by islands. To account for the argument-adjunct asymmetry exemplified by (1), it has been proposed that question particles adjacent to argument wh-in-situ can migrate to the island boundary before undergoing a second movement to Spec,CP (Hagstrom 1998), or that adjunct wh-in-situ requires a local Qu-operator whereas argument wh-in-situ does not (Aoun and Li 1993).

2 Systematic Test of the Argument-Adjunct Asymmetry

All these suggestions are based on the contrast illustrated by (1). On the basis of informal judgments, it seems that adjunct wh-in-situ in an RC island is indeed less acceptable than argument wh-in-situ.5 However, it is not so clear whether this difference in acceptability arises because different wh-phrases are sensitive to different constraints. To the best of our knowledge, no experimental study of the syntax of Chinese wh-in-situ has systematically examined the argument-adjunct asymmetry.

An account of the argument-adjunct asymmetry shown in (1) requires a consideration of various factors that could potentially control the acceptability of wh-interrogative constructions. For example, argument and adjunct wh-elements, with different functions and meanings, could exert different effects on acceptability, regardless of island effects. Furthermore, to examine whether the argument-adjunct asymmetry shown in (1) is indeed due to wh-elements moving across island boundaries, examples like (1a–b) should be compared with ones that lack island structures.

Recent experimental syntax studies have established a methodology that can be used to examine island effects systematically (Sprouse 2007, Sprouse, Wagers, and Phillips 2012, Dillon and Hornstein 2013, Sprouse and Hornstein 2013, Kush, Lohndal, and Sprouse 2018). As Sprouse, Wagers, and Phillips (2012) have shown, island effects can be characterized as superadditive interaction effects between the factor of dependency length and the presence or absence of island structures. Sentences that contain island structures are generally rated worse than sentences that do not, and longer dependencies are generally rated worse than shorter dependencies that do not span embedded clause boundaries. Sprouse, Wagers, and Phillips (2012) show that long-distance dependencies that span island boundaries result in greater acceptability decrease than what would be expected from simply adding the dependency length factor and the island structure factor; rather, long-distance extraction from within an island structure exhibits a super-additive effect (see Kush, Lohndal, and Sprouse 2018 for a detailed discussion).

To test the island effects in Chinese wh-in-situ constructions, we can employ formal experimental syntax methodology. Simply put, if Chinese wh-in-situ is sensitive to islands, we can expect superadditive interaction effects in an acceptability rating experiment. Furthermore, if adjunct wh-in-situ is more sensitive to islands, we can expect a three-way interaction of wh-element grammatical status, dependency length, and presence or absence of island structures.

3 Experiment

3.1 Experimental Design

3.1.1 Participants

Fifty-six native Mandarin Chinese speakers living in the greater Chicago area were recruited as participants.

3.1.2 Method

To test the island sensitivity of Chinese wh-in-situ, we employed a 2 × 2 × 2 factorial design with Dependency Length (short vs. long), Embedded Structure (nonisland vs. island), and Wh-Category (argument [who/what] vs. adjunct [why]) as independent factors. A sample set of stimuli with all eight conditions is shown in table 1.

Table 1

Example set of stimuli

Nonisland
ArgumentAdjunct
Short Yuehan xiangzhidao shei shuo nyuhai chi le Yuehan xiangzhidao Bier weishenme shuo 
John wonders who say girl eat ASP John wonders Bill why say 
shousi nyuhai chi le shousi 
sushi girl eat ASP sushi 
‘John wonders who said that the girl ate sushi.’ ‘John wonders why Bill says t that the girl ate sushi.’ 
Long Yuehan xiangzhidao Bier shuo nyuhai chi le Yuehan xiangzhidao Bier shuo nyuhai 
John wonders Bill say girl eat ASP John wonders Bill say girl 
shenme weishenme chi le shousi 
what why eat ASP sushi 
‘John wonders what Bill said that the girl ate.’ ‘John wonders why Bill says that the girl ate sushi t.’ 
Island
ArgumentAdjunct
Short Yuehan xiangzhidao shei jian le chi shousi Yuehan xiangzhidao Bier weishenme jian 
John wonders who meet ASP eat sushi John wonders Bill why meet 
de nyuhai le chi shousi de nyuhai 
REL girl ASP eat sushi REL girl 
‘John wonders who met the girl that ate sushi.’ ‘John wonders why Bill met the girl that ate sushi.’ 
Long Yuehan xiangzhidao Bier jian le chi Yuehan xiangzhidao Bier jian le 
John wonders Bill meet ASP eat John wonders Bill meet ASP 
shenme de nyuhai weishenme chi shousi de nyuhai 
whatREL girl why eat sushi REL girl 
‘John wonders what Bill met the girl that ate.’ ‘John wonders why Bill met the girl that ate sushi.’ 
Nonisland
ArgumentAdjunct
Short Yuehan xiangzhidao shei shuo nyuhai chi le Yuehan xiangzhidao Bier weishenme shuo 
John wonders who say girl eat ASP John wonders Bill why say 
shousi nyuhai chi le shousi 
sushi girl eat ASP sushi 
‘John wonders who said that the girl ate sushi.’ ‘John wonders why Bill says t that the girl ate sushi.’ 
Long Yuehan xiangzhidao Bier shuo nyuhai chi le Yuehan xiangzhidao Bier shuo nyuhai 
John wonders Bill say girl eat ASP John wonders Bill say girl 
shenme weishenme chi le shousi 
what why eat ASP sushi 
‘John wonders what Bill said that the girl ate.’ ‘John wonders why Bill says that the girl ate sushi t.’ 
Island
ArgumentAdjunct
Short Yuehan xiangzhidao shei jian le chi shousi Yuehan xiangzhidao Bier weishenme jian 
John wonders who meet ASP eat sushi John wonders Bill why meet 
de nyuhai le chi shousi de nyuhai 
REL girl ASP eat sushi REL girl 
‘John wonders who met the girl that ate sushi.’ ‘John wonders why Bill met the girl that ate sushi.’ 
Long Yuehan xiangzhidao Bier jian le chi Yuehan xiangzhidao Bier jian le 
John wonders Bill meet ASP eat John wonders Bill meet ASP 
shenme de nyuhai weishenme chi shousi de nyuhai 
whatREL girl why eat sushi REL girl 
‘John wonders what Bill met the girl that ate.’ ‘John wonders why Bill met the girl that ate sushi.’ 

As shown in table 1, to make the scope of wh-questions explicit and unambiguous, we constructed embedded wh-questions with embedding verbs that take an interrogative clause as their complement, such as xiangzhidao ‘wonder’. Since previous literature concerning the Chinese wh-in-situ argument-adjunct asymmetry has most widely reported and discussed data involving complex NP islands, we used complex NP islands in our island stimuli.6 There were 24 items in total, and each item tested all eight of the conditions listed above.7 All test sentences were distributed to eight lists following a Latin square design. Each list contained 24 test sentences and 72 filler sentences.8 All lists were pseudorandomized so that sentences of the same condition were not adjacent.

The test was administered using paper questionnaires. Participants were instructed to rate the naturalness of the given sentences on a scale from 1 to 7, with 1 being unnatural and 7 being perfectly natural.

3.2 Predictions

As noted above, the island effect has been shown to be characterized as a superadditive interaction effect of Dependency Length × Embedded Structure (Sprouse, Wagers, and Phillips 2012). Therefore, if wh-in-situ is island-sensitive, we would expect a superadditive interaction effect of Dependency Length × Embedded Structure. We further manipulated the category of wh-phrases (the Wh-Category factor). If, as previous studies suggest (Huang 1982a,b, Tsai 1994b, 1999), adjunct wh-in-situ exhibits greater island sensitivity than argument wh-in-situ, this greater sensitivity should be observed as a significant superadditive interaction. Thus, we would expect a significant interaction of Dependency Length × Embedded Structure × Wh-Category. On the other hand, if adjunct wh-in-situ is not more sensitive to islands than argument wh-in-situ, we would not expect such three-way interaction, and we should observe a main effect of Embedded Structure.

3.3 Results

Acceptability rating scores were recorded and analyzed using a standard ordinal regression model (Christensen 2014). Acceptability rating (1 to 7) was treated as an ordered variable predicted from the main effects and all two-way and three-way interactions of Embedded Structure, Dependency Length, and Wh-Category. Random slopes and intercepts by subject and item were included.

We found a significant main effect of Embedded Structure (β = 2.20, SE β) = 0.371, p < .001), Dependency Length (β = 3.60, SE β = 0.315, p < .001), and Wh-Category (β = 1.21, SE β = 0.281, p < .001). Sentences in island conditions received lower ratings than those in nonisland conditions; sentences with long dependencies received lower ratings than those with short dependencies; sentences with adjunct wh-in-situ received lower ratings than those with argument wh-in-situ. There was also a significant interaction of Embedded Structure × Dependency Length (β = 2.15, SE β = 0.393, p <.001). We found no significant interaction of Embedded Structure × Wh-Category (β = 0.319, SE β = 0.476, p = .503) or Dependency Length × Wh-Category (β = 0.066, SE β = 0.363, p = .855). We found no significant three-way interaction of Embedded Structure × Wh-Category × Dependency Length (β = 0.871, SE β = 0.516, p = .091).9 Mean acceptability rating scores for all conditions are shown in figure 1. Error bars represent standard errors and are calculated over aggregated data.

Figure 1

Mean acceptability rating scores

Figure 1

Mean acceptability rating scores

To study island sensitivity of each wh-category, we then nested Embedded Structure × Dependency Length within the levels of Wh-Category. We found a significant interaction of Embedded Structure × Dependency Length for both adjunct wh-in-situ (β = 1.81, SE β = 0.286, p < .001) and argument wh-in-situ (β = 2.47, SE β = 0.291, p < .001). As shown in figure 2, for both wh-in-situ categories, island conditions exhibit a greater degradation in acceptability compared with nonisland conditions when Dependency Length changes from short to long.

Figure 2

Interaction plot for adjunct and argument wh-categories separately

Figure 2

Interaction plot for adjunct and argument wh-categories separately

4 Discussion

This study offers several findings. First, we observed a main effect of Dependency Length, where a wh-in-situ located farther away from the scope position results in a lower acceptability rating. This is consistent with findings in the sentence-processing literature, where Chinese instances of wh-in-situ are found to be processed similarly to filler-gap dependencies (Xiang et al. 2014). Second, both adjunct and argument wh-in-situ show a significant interaction of Embedded Structure × Dependency Length. That is, argument wh-in-situ exhibits the RC island effect, just like adjunct wh-in-situ.10 Third, adjuncts are not more sensitive to islands than arguments, and the cause of the observed argument-adjunct asymmetry shown in (1) is a main effect of Wh-Category.

Theories that distinguish adjuncts from arguments in terms of how the operator-variable pair is formed (Huang 1982a,b, Lasnik and Saito 1984, 1992) predict a difference in island sensitivity between wh-adverbs and wh-DPs. If we adopt the concept of island effect as a superadditive interaction of Embedded Structure × Dependency Length, as proposed by Sprouse and Hornstein (2013), the difference in island sensitivity should manifest itself in a significant three-way interaction of Embedded Structure × Wh-Category × Dependency Length. More importantly, a larger Embedded Structure × Dependency Length interaction is predicted for wh-adverbs than for wh-DPs. However, no significant three-way interaction was found. Although the p-value for the three-way interaction is arguably small and might indicate marginal significance ( p = .091), we should note that the direction of the effect underlying this marginal significance is numerically opposite from what Huang’s (1982a,b) and Lasnik and Saito’s (1984, 1992) proposals would predict. The Embedded Structure × Wh-Category interaction is larger for wh-DPs. In addition, the significant main effect of Wh-Category indicates that wh-adverbs are indeed worse than wh-DPs in the complex NP island context, but the reason is not a higher island sensitivity. Rather, wh-adverbs are judged worse than wh-DPs across the board, whether within an island or not.

Furthermore, theories that suppose that argument wh-in-situ does not move (Tsai 1994a,b, 1999) predict no interaction of Dependency Length × Embedded Structure for wh-DPs. However, like the adverb conditions, the DP conditions exhibited a significant interaction of Dependency Length × Embedded Structure (see figure 2); that is, the island structure exhibited a superadditive effect. Other theories proposing the movement of a question particle (or feature) in the case of wh-in-situ would also be challenged: with no difference in island sensitivity across wh-categories, the claims that question particles adjacent to argument wh-in-situ can avoid island constraints by migrating to island boundaries (Hagstrom 1998) or by originating outside islands (Aoun and Li 1993) would also be unsupported.

Our findings challenge the long-standing generalization that wh-adverbs in Chinese are more sensitive to islands than wh-DPs, and thus challenge the aforementioned theories that capture such “argument-adjunct asymmetry” in terms of island sensitivity. In other words, without the argument-adjunct asymmetry, we can potentially discard the stipulation that the Subjacency Condition only constrains overt movements (Huang 1982a,b) or that wh-DPs do not undergo covert movement while wh-adverbs do (Tsai 1994a,b, 1999). If subjacency effects are explained by phrasal movement, our results should also suggest that both wh-DPs and wh-adverbs in Chinese undergo phrasal movement.11 In more modern frameworks of syntax, islands can be understood as Spell-Out domains (Uriagereka 1999, Nunes and Uriagereka 2000, Chomsky 2008); once a domain is spelled out, the elements it contains are not accessible for further syntactic operations. In such a framework, our results suggest that both argument and adjunct wh-elements are inaccessible after Spell-Out.

Our study raises an intriguing question: what led researchers to the generalization that an argument-adjunct asymmetry exists in Chinese wh-in-situ island sensitivity? It is important to note that the original observations by Huang (1982a,b) and Tsai (1994a,b) are by no means incorrect. Examples (1a) and (1b), cited from these works, reflect the island/argument/long condition and the island/adjunct/ long condition in our design, respectively. As shown in figure 1, the island/argument/long condition indeed has a higher acceptability rating than the island/adjunct/long condition. Thus, the argument-adjunct difference reported in (1) does exist. However, basing the claim that only adjunct wh-in-situ is sensitive to complex NPs solely on this data point is potentially problematic, as this observation can very well be reflecting a main effect of Wh-Category rather than the adjunct’s extra sensitivity to islands. Acceptability judgments of minimal pairs can point to the relevant main effects, but such methodologies are not designed to detect two-way or three-way interactions of factors. Thus, we speculate that the specific design of the judgment-of-minimal-pairs method led to the decades-long generalization regarding the argument-adjunct asymmetry in terms of island sensitivity.

Another question our results raise is why weishenme ‘why’ is judged worse than argument wh-in-situ in general. There are many possible reasons, but we tentatively suggest that it is because weishenme has multiple meanings (‘reason why’ and ‘purpose why’; see Stepanov and Tsai 2008 for details) and because resolving this ambiguity might cause some difficulty. This point clearly requires more systematic investigation.

The current study is limited to complex NP islands. To make a more generalized argument for or against argument-adjunct asymmetry, we should test other island types (e.g., adjunct islands, subject islands), adopting similar methodology. In addition, to get a more general picture of the syntax of wh-in-situ, similar experimental investigations should be extended to wh-in-situ in other contexts (e.g., multiple wh-questions). There are several approaches to wh-in-situ in multiple wh-questions: the covert phrasal movement analysis (Karttunen 1977, Huang 1982a,b, Hornstein 1995, Pesetsky 2000), the wh-feature movement analysis (Chomsky 1995, Lasnik 1995, Ochi 2015), and interpretation of wh-in-situ in its original position (Baker 1970, Aoun and Li 1993, Reinhart 1997, Cheng 2009). If wh-in-situ in English (and other languages) exhibits island sensitivity similar to that demonstrated here for Chinese, the covert phrasal movement approach would be supported.

In addition, regarding crosslinguistic variations of wh-in-situ, we consider investigating wh-islands an important next step. It has been suggested that wh-in-situ in Chinese and wh-in-situ in Korean/Japanese differ in terms of wh-island sensitivity (Watanabe 1992, Aoun and Li 1993). Sensitivity to wh-islands has been taken as a key property to distinguish the two types of wh-in-situ. The experimental methodology employed in the present study can be particularly useful to examine whether or not instances of wh-in-situ in these languages show asymmetry in wh-island sensitivity.

5 Conclusion

In this study, we found that both adjunct and argument wh-in-situ in Chinese are sensitive to complex NP islands and that there is no difference in island sensitivity between the two types of wh-in-situ. A main effect of Wh-Category (adjunct vs. argument) underlies the argumentadjunct asymmetry shown in (1). Therefore, this study supports the covert (LF) phrasal movement analysis of wh-in-situ in Chinese.

Notes

1 In this study, we restrict our attention to Chinese wh-in-situ, and we do not investigate wh-in-situ in other languages or wh-in-situ in multiple wh-questions (see Cheng 2009 for a survey of the syntax of wh-in-situ in a broader context). This limitation is due mainly to the lack of experimental data for wh-in-situ in other contexts, within Chinese or elsewhere; as a result, we can develop experimental conclusions only about Chinese wh-in-situ of the type illustrated in (1). We will discuss broader issues with respect to wh-in-situ in section 3.

2 To keep the terminology theory-neutral, we call such islands relative clause (RC) islands.

3 In previous literature, the island insensitivity of Chinese argument wh-in-situ has mostly been presented in the context of RC islands as in (1) (Huang 1982a,b, Tsai 1994a,b, Cheng 2009). Other studies have also presented or cited examples with complex NP islands as important data points among examples with other island types (Watanabe 1992, Aoun and Li 1993). Thus, we cite (1a–b) to illustrate previous arguments for asymmetry in the island sensitivity of Chinese adjunct/argument wh-in-situ.

4 Notably, Lasnik and Saito’s (1984, 1992) explanation includes the possibility of deleting traces in LF, which is crucial to explaining the argument-adjunct asymmetry shown in (1). Because this squib does not address the ECP per se, we do not explore the issue of trace deletion any further.

5 Note that we are not claiming that argument wh-in-situ is perfectly acceptable within RC islands. Cheng (2009:769) assigns “??” to (i), an example of argument wh-in-situ in an English multiple wh-question context, suggesting that argument wh-in-situ in a complex NP island is not perfectly acceptable (cf. results reported by Sprouse et al. (2011), who did not detect a complex-NP island effect in English multiple wh-questions).

The same could hold true for Chinese wh-in-situ, even though native speakers of Chinese judge examples like (1a) as quite acceptable and detect a clear difference between (1a) and (1b). The most important point for us is that argument wh-in-situ is judged as better than adjunct wh-in-situ within the island context.

6 Huang (1982a,b), Tsai (1994a,b), and Cheng (2009) all present and discuss Chinese argument-adjunct asymmetry data involving complex NP islands. Aoun and Li (1993) cite data involving subject islands, adjunct islands, and complex NP islands. Watanabe (1992) presents data involving wh-islands and complex NP islands.

7 In the experiment, all the stimuli were shown in Chinese characters. Here, for the sake of readability, we show the romanized examples.

8 Filler items were sentences chosen to be apparently acceptable or unacceptable in Chinese. The 36 acceptable sentences included simple SVO sentences joined by conjunctions, subject and relative clause constructions, cleft and pseudocleft constructions, and double object constructions. The 36 unacceptable sentences included the same constructions, but all copulas were replaced by aspect markers, and aspect markers by copulas. Filler items were matched with experimental items and each other in terms of average length and complexity (number of verbs per sentence).

9 We decided not to do a power analysis, for the following reasons: (1) It is potentially problematic to calculate power for general mixed-effect regression models given the uncertain degree of freedom. (2) Even if we were to fail to detect a three-way interaction due to the lack of power, the fact that wh-DP island sensitivity is numerically greater than wh-adverb island sensitivity should still make a compelling argument that we found no evidence for the original generalization of argument-adjunct asymmetry.

10 An anonymous reviewer noted that the acceptability degradation observed in nonisland conditions was possibly caused by interpretive ambiguity in the nonisland/adjunct conditions, not by island sensitivity. To check whether such an interpretive ambiguity exists, we asked five native speakers of Chinese whether they could detect ambiguity in any of the sentences used in our experiment. According to these speakers, in the nonisland/adjunct/long condition the word weishenme ‘why’ unambiguously modifies the embedded clause verb; in the nonisland/adjunct/short condition, the word weishenme ‘why’ unambiguously modifies the matrix verb. Thus, we are confident that no interpretive ambiguity is involved in the nonisland/adjunct conditions.

11 By covert movement, we mean that in Chinese both wh-DPs and wh-adverbs undergo a similar type of movement and that they do not go through different processes of forming operator-variable pairs (as in phrasal movement vs. feature movement). Multiple analyses of movement are possible, including copy theories (Chomsky 1995, Nunes 2001, 2004) and multidominance theories (Epstein et al. 1998, Gärtner 2002, Johnson 2012). In this study, however, we are not making claims favoring any specific analysis of movement.

Acknowledgments

This work was partially supported by NIH-NIDCD R01 DC01948 grant (PI: Cynthia K. Thompson) and Northwestern University Undergraduate Research Grant (awarded to Jiayi Lu). We are grateful to the two anonymous reviewers, the audiences at AMLaP 2018 and LSA 2019, Norbert Hornstein, Shayne Sloggett, and Ming Xiang for their valuable comments and suggestions. We would also like to thank Xunru Che and Yilan Wang for their Mandarin Chinese judgments and help with item creation.

References

References
Aoun,
Joseph
, and
Yen-hui Audrey
Li
.
1993
.
Wh-elements in situ: Syntax or LF?
Linguistic Inquiry
24
:
199
238
.
Baker,
C. L.
1970
.
Notes on the description of English questions
.
Foundations of Language
6
:
197
219
.
Cheng,
Lisa Lai-Shen
.
2009
.
Wh-in-situ, from the 1980s to now
.
Language and Linguistics Compass
3
:
767
791
.
Chomsky,
Noam
.
1977
. On wh-movement. In
Formal syntax
, ed. by
Peter
Culicover
,
Thomas
Wasow
, and
Adrian
Akmajian
,
71
132
.
New York
:
Academic Press
.
Chomsky,
Noam
.
1981
.
Lectures on government and binding
.
Dordrecht
:
Foris
.
Chomsky,
Noam
.
1986
.
Barriers
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
.
1995
.
The Minimalist Program
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
.
2008
. On phases. In
Foundational issues in linguistic theory
, ed. by
Robert
Freidin
,
Carlos P.
Otero
, and
Maria Luisa
Zubizarreta
,
133
166
.
Cambridge, MA
:
MIT Press
.
Christensen,
Rune Haubo Bojesen
.
2014
.
Analysis of ordinal data with cumulative link models—estimation with the R-package ordinal
. https://cran.r-project.org/web/packages/ordinal.
Dillon,
Brian
, and
Norbert
Hornstein
.
2013
. On the structural nature of island constraints. In
Experimental syntax and island effects
, ed. by
Jon
Sprouse
and
Norbert
Hornstein
,
208
222
.
Cambridge
:
Cambridge University Press
.
Epstein,
Samuel David
,
Erich M.
Groat
,
Ruriko
Kawashima
, and
Hisatsugu
Kitahara
.
1998
.
A derivational approach to syntactic relations
.
New York
:
Oxford University Press
.
Gärtner,
Hans-Martin
.
2002
.
Generalized transformations and beyond: Reflections on Minimalist syntax
.
Berlin
:
Akademie Verlag
.
Hagstrom,
Paul
.
1998
.
Decomposing questions
.
Doctoral dissertation, MIT, Cambridge, MA
.
Hornstein,
Norbert
.
1995
.
Logical Form: From GB to Minimalism
.
Cambridge, MA
:
Blackwell
.
Huang,
C.-T. James
.
1982a
.
Logical relations in Chinese and the theory of grammar
.
Doctoral dissertation, MIT, Cambridge, MA
.
Huang,
C.-T. James
.
1982b
.
Move wh in a language without wh-movement
.
The Linguistic Review
1
:
369
416
.
Johnson,
Kyle
.
2012
.
Towards deriving differences in how wh movement and QR are pronounced
.
Lingua
122
:
529
553
.
Karttunen,
Lauri
.
1977
.
Syntax and semantics of questions
.
Linguistics and Philosophy
1
:
3
44
.
Kush,
Dave
,
Terje
Lohndal
, and
Jon
Sprouse
.
2018
.
Investigating variation in island effects
.
Natural Language and Linguistic Theory
36
:
743
779
.
Lasnik,
Howard
.
1995
. Verbal morphology: Syntactic Structures meets the Minimalist Program. In
Evolution and revolution in linguistic theory: Studies in honor of Carlos P. Otero
,
251
275
.
Washington, DC
:
Georgetown University Press
.
Lasnik,
Howard
, and
Mamoru
Saito
.
1984
.
On the nature of proper government
.
Linguistic Inquiry
15
:
235
289
.
Lasnik,
Howard
, and
Mamoru
Saito
.
1992
.
Move α: Conditions on its application and output
.
Cambridge, MA
:
MIT Press
.
Nunes,
Jairo
.
2001
.
Sideward movement
.
Linguistic Inquiry
32
:
303
344
.
Nunes,
Jairo
.
2004
.
Linearization of chains and sideward movement
.
Cambridge, MA
:
MIT Press
.
Nunes,
Jairo
, and
Juan
Uriagereka
.
2000
.
Cyclicity and extraction domains
.
Syntax
3
:
20
43
.
Ochi,
Masao
.
2015
. Wh-adjuncts, left periphery, and wh-in-situ. In
Chinese syntax in a cross-linguistic perspective
,
401
428
.
New York
:
Oxford University Press
.
Pesetsky,
David
.
2000
.
Phrasal movement and its kin
.
Cambridge, MA
:
MIT Press
.
Reinhart,
Tanya
.
1997
.
Quantifier scope: How labor is divided between QR and choice functions
.
Linguistics and Philosophy
20
:
335
397
.
Ross,
John Robert
.
1967
.
Constraints on variables in syntax
.
Doctoral dissertation, MIT, Cambridge, MA
.
Sprouse,
Jon
.
2007
.
A program for experimental syntax: Finding the relationship between acceptability and grammatical knowledge
.
Doctoral dissertation, University of Maryland, College Park
.
Sprouse,
Jon
,
Shin
Fukuda
,
Hajime
Ono
, and
Robert
Kluender
.
2011
.
Reverse island effects and the backward search for a licensor in multiple wh-questions
.
Syntax
14
:
179
203
.
Sprouse,
Jon
, and
Norbert
Hornstein
.
2013
.
Experimental syntax and island effects
.
Cambridge
:
Cambridge University Press
.
Sprouse,
Jon
,
Matt
Wagers
, and
Colin
Phillips
.
2012
.
A test of the relation between working memory capacity and syntactic island effects
.
Language
88
:
82
123
.
Stepanov,
Arthur
, and
Wei-Tien Dylan
Tsai
.
2008
.
Cartography and licensing of wh-adjuncts: A cross-linguistic perspective
.
Natural Language and Linguistic Theory
26
:
589
638
.
Tsai,
Wei-Tien Dylan
.
1994a
.
On economizing the theory of A-bar dependencies
.
Doctoral dissertation, MIT, Cambridge, MA
.
Tsai,
Wei-Tien Dylan
.
1994b
.
On nominal islands and LF extraction in Chinese
.
Natural Language and Linguistic Theory
12
:
121
175
.
Tsai,
Wei-Tien Dylan
.
1999
.
On lexical courtesy
.
Journal of East Asian Linguistics
8
:
39
73
.
Uriagereka,
Juan
.
1999
. Multiple spell-out. In
Working Minimalism
, ed. by
Samuel David
Epstein
and
Norbert
Hornstein
,
251
282
.
Cambridge, MA
:
MIT Press
.
Watanabe,
Akira
.
1992
.
Wh-in-situ, subjacency and chain formation
.
MIT Occasional Papers in Linguistics 2
.
Cambridge, MA
:
MIT, MIT Working Papers in Linguistics
.
Xiang,
Ming
,
Brian
Dillon
,
Matt
Wagers
,
Fengqin
Liu
, and
Taomei
Guo
.
2014
.
Processing covert dependencies: An SAT study on Mandarin wh-in-situ questions
.
Journal of East Asian Linguistics
23
:
207
232
.
Xu,
Liejiong
.
1990
.
Remarks on LF movement in Chinese questions
.
Linguistics
28
:
355
382
.