Control and raising passives, and why Mandarin does not smuggle

Collins (Syntax 8:81–120, 2005) proposes a smuggling approach to English passives that solves some problems associated with the traditional analysis. This article will show that while English passives involve smuggling, Mandarin passives do not; we offer an explanation for this difference. We first provide evidence that Mandarin passives can have not only control structures (as previously assumed) but also the possibility of a raising derivation involving A-movement, thus ruling out control as the sole reason for the absence of smuggling. We then attribute the absence of smuggling in Mandarin to the existence of vP-internal movement, which implies that Chinese allows multiple Specs of vP while English does not. This analysis helps tie together a number of otherwise unrelated differences between these languages (vP-internal movement, quantifier float, and constituency). We see the results as falling within the basic tenets of the theory of UG: While UTAH and Minimality are presumably universal requirements, individual grammars may employ different strategies to satisfy them.

1 Introduction Collins (2005) points out that the standard analysis of the English be passive (Jaeggli 1986;Baker 1988;Baker et al. 1989, and many others), with the external argument (EA) appearing in different syntactic positions for actives and passives, violates Baker's (1988, p. 46) Uniformity of Theta-Assignment Hypothesis (UTAH). He proposes an underlying structure like (1), in accordance with UTAH, with the EA appearing in Spec of vP for the passive as it does in the active: (1) [ TP e T [ VP be [ VoiceP by [ vP John [ v'

v [ PartP written the book]]]]]]
In order to form the passive, the underlying object the book needs to move to e in the Spec of TP. However, direct movement of the book to e crossing the EA John is prevented by Relativized Minimality (RM) or the Minimal Link Condition (MLC). To circumvent this problem, Collins (2005, p. 90) proposes a smuggling approach to passivization, as illustrated in (2): (2) The main points of this hypothesis are as follows. By is the head of VoiceP taking a vP as its complement. The external argument is merged in Spec, vP in the same way as in the active. Movement of the internal argument the book to the Spec, TP position is carried out in 2 steps: First, the participle phrase written the book is moved to Spec, VoiceP, and then the book is moved to Spec, TP. The two-step process effectively smuggles the internal argument to Spec, TP crossing the EA John without violating RM or the MLC. The result strands the EA in Spec, vP but, crucially, does not produce an adjunct PP as in traditional analyses. The underlying structure of be passives proposed by Collins comes close to that of Mandarin bei passives proposed by Huang (1999) and Huang, Li & Li (2009, henceforth HLL), in which the external argument is the subject of the complement clause of the passive marker bei. A major difference occurs in surface word order between English and Mandarin, however: In English, the main (participle) VP appears before the agent, but in Mandarin, the VP follows the agent-precisely as it does in the corresponding active. This is clear from the 'long-passive' example in (3) below with an overt external argument (Lisi). Although this is less clear from the 'short (agentless) passive' in (4), everyone who postulates an implicit agent for it will no doubt place it before the VP as well: (3) Mandarin long passive: Subjectbei -EA -VP Zhangsan bei Lisi da le. Zhangsan BEI Lisi hit LE 1 'Zhangsan was hit by Lisi.' (4) Mandarin short (agentless) passive: Subjectbei -[EA] -VP Zhangsan bei da le. Zhangsan BEI hit LE 'Zhangsan was hit.' There are two logical possibilities to look at this word-order difference between English and Chinese. The first is to simply take it to mean that while in English an object needs to be smuggled in a VP across the Agent phrase before moving to the surface subject position, in Chinese the derivation seems more simple, as the object can move by itself without smuggling. Under this hypothesis, the question then arises how this is possible, assuming (as is natural and independently supported) all the relevant conditions (UTAH, RM, and MLC, etc.) to be applicable to both languages. The second possibility is that derivation of the Chinese cases is more complex than meets the eye, as it in fact involves smuggling but its word-order effect becomes invisible for some reason. If we take this hypothesis, it remains to spell out how smuggling is done and its effect gets ''undone'' on the surface.
In this paper, we argue for the first hypothesis and propose a non-smuggling approach for Mandarin passives and relate the word-order difference to other independent differences between the two languages. This will be carried out in Sect. 3, where we also briefly take up the second hypothesis (and argue against it). Before we get to the main problem, however, we need to be clear of where the problem is. We address this matter first in Sect. 2.
2 Reanalysis of the structure of Mandarin bei passives 2.1 The control/predication analyses of Mandarin bei passives According to Feng (1995), Ting (1995), Huang (1999) and HLL (2009) among others, the Mandarin long bei passive can be analyzed as a structure of complex predication, as shown in (5).
Zhangsan BEI Lisi hit LE 'Zhangsan was hit by Lisi.' In (5), the semi-lexical verb bei selects an active IP complement, within which a null operator (NOP) moves from the object position of the base verb and adjoins itself to IP. Bei and the NOP-clause form a complex predicate, which selects Zhangsan as its subject argument. The moved NOP object is coindexed with the matrix subject under predication.
The structure of the short bei passive is a control structure (Huang 1999;HLL 2009): Lisi BEI hit-LE 'Lisi was hit' In (6), bei selects an Experiencer subject and a VP complement whose Theme object (a PRO) is moved to Spec, VP, where it is controlled by the matrix subject.
The control/predication analysis for the bei passive receives considerable support from a number of facts. First, the subject of bei may take subject-oriented adverbs (e.g., guyi 'intentionally'), justifying a base-generated position in which it is assigned an Experiencer theta-role: (7) Zhangsan guyi bei (Lisi) da le. (Huang 1999, p. 5) Zhangsan intentionally BEI Lisi hit LE 'Zhangsan intentionally got hit by Lisi.' Secondly, the long passives exhibit A'-movement properties, such as longdistance dependencies, as in (8), island effects and possible occurrence with resumptive pronouns as in (9), and occurrence of the relativization marker suo, as in (10).  (HLL 2009, p. 125) (10) zhexie shiqing bu neng bei tamen suo liaojie. (HLL 2009, p. 126) these thing not can BEI they SUO understand 'These things cannot be understood by them.' The NOP-movement analysis depicted in (5) assimilates the long passive to a tough construction (as analysed by Chomsky 1981). Like the tough construction, it exhibits long-distance dependencies as in (8), subject to island constraints as in (9). The fact that the derivation can be saved with a resumptive pronoun as in (9) further supports the A'-movement hypothesis because such a pronoun always appears in an A'-bound position. As for (10), Chiu (1995) has shown that the long passive may include the particle suo in the immediate pre-verbal position. In the mean time, it has been well known that the only other construction with suo is the relative clause construction, a typical A'-movement structure, suggesting that the long passive itself has a similar structure. 2 Now, if all passives are derived as described, via A'-movement followed by Predication as in (5) or via short A-movement followed by control as in (6), then there is already an answer to the question why Chinese passives do not involve smuggling of the object. In (5), A'-movement of the NOP across the EA Lisi is no problem since the latter is in an A-position. In (6), short A-movement around VP does not cross the EA position either, because as assumed in Huang (1999), the EA has been suppressed and does not occur in the structure. In neither case does the movement violate RM or MLC, and smuggling is entirely unnecessary (hence also prohibited).
However, in more recent works, Huang (2013) and Liu (2012) argued for the possibility of a raising derivation for some passive sentences in Chinese. If their claim is correct, then the question pops up again why Chinese does not resort to smuggling. We briefly summarize their reasoning and add a new argument for the possibility of raising below.

2.2
The possibility of a raising analysis for bei passives Huang (2013) and Liu (2012) observe that short passives and local long passives (i.e., the long passives that do not exhibit long distance dependencies) allow idiomchunks to be fronted under passivization, as in (11)(12). Such examples imply a raising analysis for the idiom-chunk subject pianyi 'advantage': (11) pianyi dou bei (ta yi-ge-ren) zhan-guang-le advantage all BEI (he alone) take-empty-LE 'All the advantage was taken (by him single-handedly).' (12) zhe-zhong mo yijing bei you-guo haoji ci le. this-kind -mor already BEI hu-Exp several time LE 'This type of saying humorous things has been done several times already.' (Lit.: you mo = humor 'This -mor has been hu-ed several times already.') Such cases do not arise in a sentence with a subject-oriented adverb like guyi 'intentionally'. This situation is as predicted, because such an adverb implies subject thematicity, and idiom-chunks do not bear argument roles. Given these considerations, when a passive involves neither subject-oriented adverbs nor idiom chunks-as in (13), logically either a raising or a control analysis is possible: his friend BEI Lisi criticize-LE 'His friend got criticized (by Lisi).' There is an additional argument for the possibility of a raising analysis that we can offer. As has been well known in the literature, in Chinese a passive sentence typically describes an event understood as being unfortunate for or adversative to some individual or other. Very often the individual concerned, being physically or psychologically affected, is the subject of bei. This situation is aptly captured in the control/predication analysis, according to which the subject receives an independent theta role, Experiencer/Affectee, which serves as the antecedent of the PRO or NOP under predication. However, it has also been well known that the experiencer/ affectee of a passive event is not always the subject of bei. The experiencer/affectee can be expressed as an oblique, applicative argument, or its existence may be simply implied. This is clearly the case with an idiom-chunk subject. In the case of (14), the subject cannot possibly be an Experiencer, yet the event of all advantage being taken by him alone may be understood as being unfortunate to someone, possibly the speaker or some other salient individual in context. In fact, the experiencer/ affectee may appear as an oblique argument as in (14): (14) pianyi dou bei ta gei wo zhan-guang-le! advantage all BEI he on me take-empty-LE 'All the advantage was taken by him on me!' In other words, when the subject is not the psychological experiencer, another argument (implicit or explicit) bearing the role Experiencer still exists. The same is true with a non-idiom-chunk subject as well, as shown in (15a), where an implicit experiencer may be felt to exist (e.g., the speaker or the shoes' owner) and (15b), where the experiencer is expressed.
(15) a. na-shuang xiezi bei ta ti-po-le that-CL shoes BEI he kick-broken-LE 'That pair of shoes were kicked-broken by him.' b. na-shuang xiezi bei ta gei wo ti-po-le that-CL shoes BEI him on me kick-broken-LE 'That pair of shoes were kicked-broken by him on me.' A corresponding example is provided from Taiwanese Southern Min (TSM) with an optional Affectee phrase ka gua 'on me': (16) hit-shiang e-a hoo yi (ka gua) that-pkua-khi a. that-CL shoes BEI him on me kick-broken-away LE 'That pair of shoes were kicked-broken by him (on me).' The existence of a (possibly implicit) experiencer that is distinct from the subject is entirely natural in each case, since an inanimate subject cannot be a (psychological) Experiencer by definition, but may be a Theme/Patient. With an animate subject, a sentence may be ambiguous depending on whether the subject is an Experiencer and a pure Theme/Patient. As a concrete example, consider (17): 3 (17) xiao haizi bei ta da-de bi-qing-lian-zhong le.
small child BEI he hit-DE black-and-blue LE 'The little child was hit black-and-blue by him.' The subject 'small child' may be the experiencer (the sufferer) of an event in which he/she is the Theme object, or the child may simply be the Theme of the event while someone else (e.g., the speaker, who might be the child's mother) is the mental experiencer/sufferer. Under the latter interpretation, the 'someone else' Experiencer can be optionally expressed as in (18): (18) xiao haizi bei ta (gei wo) da-de small child BEI he (on me) hit-DE bi-qing-lian-zhong le. black-and-blue LE 'The little child was hit black-and-blue by him (on me).' This state of affairs strongly suggests that the semi-lexical passive verb bei fluctuates between a control and a raising verb. The raising verb is derived from the control verb as a result of argument suppression. In particular, the raising examples (15a)-(15b) can be related to their control 'source' below: (19) wo bei ta ti-po-le na-shuang xiezi. I BEI he kick-broken-LE that-CL shoes 'I had that pair of shoes kicked-broken by him.' (i) na-feng xin juran bei ta pai ren tou-zou-le! that-CL letter to-my-dismay BEI him send person steal-away-LE 'That letter (to my dismay) got ''sent-people-to-steal-away'' by him!' This may imply that the inanimate subject ne-feng xin 'that letter' has arrived at its matrix subject position via movement. We think this is possible given a recent analysis of the tough-construction by Hartman (2012). Based on certain intervention effects with matrix experiencers in tough constructions, Hartman argues for a two-step analysis with ''A'-movement to the edge of the highest embedded clause, followed by A-movement to the matrix subject position'' (2012, p. 97). Following Hartman's line raises other issues, however, which we shall not pursue here.
Control and raising passives, and why Mandarin does not smuggle 391 In standard GB terms, suppression of the Experiencer subject in (19) will trigger A-movement of the object to the subject position. The Experiencer subject remains implicit (and existentially closed) as in (15a), or appears as an applicative geiphrase, as in (15b). This is parallel to a familiar assumption about standard bepassives in English: passive morphology suppresses the EA, the object is moved to Spec, TP, and the EA may remain, either as an implicit argument or expressed in a by-phrase. There is also an (almost) exact parallel here with the two uses of certain/sure: (20) a. John is certain that he will win.
b. John is certain [t to win].
Under coreference between John and he, (20a) is akin to a control sentence (cf. John is eager to win). Suppression of the higher subject argument leads to the raising structure (20b), where the Experiencer is either implicit (often the speaker) or expressed (e.g., to me). Similarly, familiar raising verbs like seem and appear involve an implicit Experiencer (cf. Middle English methinks 'it seems to me'). 4 In short, an argument for allowing a raising analysis for certain Mandarin bei passives is that it enables one to identify the bearer of a misfortune not otherwise expressed in the sentence as an implicit Experiencer. The conclusion that raising is possible means, of course, that we have to face the question why Mandarin raising passives do not seem to smuggle. In the next section we shall entertain new structures for Mandarin local long passives.
3 The new structures of Mandarin local long bei passives

No smuggling occurs in the Mandarin bei passive
We propose that the absence of smuggling in Mandarin is related to another property of Mandarin, i.e., Mandarin allows an object to be preposed to a postsubject but pre-verbal position (Ernst and Wang 1995;Shyu 1995), as illustrated in (21). However, such movement is not available in English.
(i) xiao-niao gei fei-zou le. small-bird give fly-away LE 'The little bird flew away (on someone).' Huang argues that gei 'give' here is actually an existential raising verb 'happen' that is derived from (ii) by subject suppression and subsequent raising of xiao-niao 'little bird'. The suppressed subject remains in (i) as an implicit Experiencer, existentially closed, in a way parallel to cases discussed in the text.
(21) a. Lisi kan-guo le na-ben shu (Shyu 1995, p. 100) Lisi read-Exp LE that-CL book 'Lisi has read that book.' b. Lisi na-ben shu i kan-guo le t i (object preposing structure) Lisi that-CL book read-Exp LE 'Lisi has read that book.' Ernst and Wang (1995) argue that the preposed object is adjoined to VP. Gaining insights from their analysis and following Richards (1997) and Ura (1996Ura ( , 2000, we assume that the head v allows two specifiers, and that an object, such as na-ben shu 'that book' in (21b), may move to the inner Spec, vP. The external argument Lisi is merged in the outer Spec, vP and moves to Spec, TP to check its Case feature and satisfy the EPP feature of T.
In addition, Shyu (1995, p. 105ff) argues that object preposing as in (21b) (23) and (24) show that object preposing lacks obligatory Binding Principles A and C reconstruction effects. (25) shows that object preposing can rescue a sentence from weak crossover effects. Based on these facts, we deduce that the inner Spec, vP position is an A-position. 5 Based on the availability of object preposing in Mandarin but not in English, we assume that Chinese allows multiple Specs of vP while English allows a single Spec of vP. It should be noted that the Spec of vP mentioned here is different from the edge of v*P in Chomsky's (2004Chomsky's ( , 2008 sense. According to Chomsky (2008), v*P is a phase and the outer Spec, v* is an A'-position. For instance, in the derivation of the sentence Who did John see?, who moves to the outer Spec, v* and goes on moving to Spec, CP, as in ]. The edge of v*P is available language-universally. However, the multiple Specs of vP mentioned here are only available in a language like Chinese but not English-like languages. As for why English and Chinese differ with respect to the availability of multiple vP specifiers, we simply take this as reflecting a parametric difference, given the possible occurrence of an overt preposed object in Chinese (and possibly also clause-internal scrambling in Japanese, etc.), and its impossibility in English. This assumed parametric difference about vP is, in fact, parallel to one proposed by Richards (1997) concerning the availability of multiple CP specifiers. Richards (1997) argues that, by assuming that Chinese (together with Japanese, Bulgarian, etc.) have multiple CP specifiers (which enable 'tucking-in' operations) while English-type languages do not, a number of well known differential behaviors of wh-movement between these two types of languages can be made to fall out nicely. To the extent that Richard's (1997) CP-Spec parameter is justified, our vP-Spec parameter can be seen as a welcome natural extension from CP to vP, i.e., a generalization of the 'tucking-in parameter' to both phase domains.
We further assume that the multiple Specs of vP can be an escape hatch in Mandarin passives for the object moving on its way to Spec, TP without violating minimality conditions. English, however, does not have such an escape hatch for the object. In order to avoid the violation of minimality conditions, smuggling is therefore called for.

The derivations of Mandarin local long bei passives
Some basic assumptions are made as follows: First, as Collins (2005) assumes, the preposition by in the English be passive occupies the head of VoiceP because the distribution of agentive by-phrase is restricted to passive. Similar to by, the Mandarin passive marker bei only occurs in the passive and with the external argument. Therefore, we assume that Mandarin passives involve a VoiceP headed by bei, which takes a vP complement. The Spec of vP is occupied by the external argument.
Moreover, according to Huang's (2013) proposal of ''Passivization Cartography'', the fluctuating properties of Mandarin bei passives (between control and raising) may be caused by the semi-lexical verb bei occupying more than one point on the causative-unaccusative spectrum (cause [ let [ witness [ undergo [ be affected by [ become [ exist [ be). We hence suppose that bei can be decomposed into EXPERIENCE (Exp) and BECOME (Bec) components in local long bei passives. 6 The raising structure of a bei passive involves only the BECOME component, and the derivation of (3) is shown in (26): After the external argument Lisi is merged in Spec, vP, the internal argument Zhangsan of the main verb da 'hit' moves to the Spec of vP below the external argument in the manner of ''tucking in'' (cf. Richards 1997) since vP in Chinese allows multiple Specs. V moves to v. Importantly, according to Chomsky (1995:184-185), the internal and external arguments are equidistant to higher destinations of movement. The Voice head bei is merged with vP, checks its accusative Case feature with the external argument and makes it inactive. Therefore, only the internal argument remains active and is available for movement. The head Bec is merged and the morpheme bei moves to adjoin to it. The internal argument moves to Spec, VoiceP and then to Spec, BecP and Spec, TP to check Nominative Case. There is no violation of minimality and hence smuggling is not needed.
The control structure of the local long passive involves both the Experience (Exp) and Become (Bec) components of bei. 7 The derivation of (3) is shown as in (27), in which the passive verb bei is base-merged in Voice and then moves to Bec and Exp. The PRO object of the verb 'hit' moves to inner Spec, vP, then to Spec, VoiceP and Spec, BecP. Exp introduces the Experiencer subject Zhangsan, which controls PRO in Spec, BecP. Similar to the raising case, smuggling is not needed.
We have shown that smuggling is not necessary in the derivation of Mandarin passives. Below we present additional evidence that smuggling indeed does not take place in Mandarin.

The quantifier floating test
Following Sportiche (1988), quantifiers and the DPs they quantify are commonly considered to originate as a single constituent. The positions where a quantifier floats are the ones through which DP movement passes. The distribution of the floated quantifiers in Mandarin bei and English be passives as in (28) (29) a. *They were arrested by the police all.
b. They were all arrested by the police.
In Mandarin bei passives, the floated quantifier quanbu 'all' can appear either below bei ? Agent DP, as in (28a), or above it, as in (28b). However, in English be passives, all cannot float below the agentive by-phrase, as in (29a). The difference between bei passives and be passives can be accounted for if we assume that Mandarin allows multiple Specs of vP but no smuggling, while English has smuggling but not multiple Specs of vP. The explanation goes as follows: The derivations of (28a, b) are illustrated in (30a, b) respectively. In (30a), the quantifier quanbu 'all' is assumed to be floated in the inner Spec, vP position. In (30b), quanbu is floated in the Spec, BecP position. This shows that on its way moving to Spec, TP, the object has passed through these positions.
However, in the derivation of the English passive (29b), as shown in (31), the PartP smuggles the object to Spec, VoiceP, a position above the Agent by-phrase. That is why all cannot float below the agentive by phrase. 8 8 Note that quantifiers cannot be floated at positions immediately following main verbs in English or Chinese: (i) a. They were all arrested by the police. b. * They were arrested all by the police.

The ''by-phrase'' constituency test
According to Huang (1999) and HLL (2009), unlike the by-phrase in English, the bei-DP in Mandarin does not behave as a constituent; i.e., it cannot move across a time phrase or prepose to a sentence initial position, as shown in (32b, c) respectively: Footnote 8 continued This is the case even if the verbal head is followed by another predicate: (iii) a. John caused them all to leave early.
b. *They were caused all to leave early.
Hence it is also impossible to float a quantifier after bei, even when the latter has raised to Bec. The following is ungrammatical with quanbu stranded at Spec, VoiceP following the raised bei: (iv) *pingguo bei quanbu Lisi mai-zou-le (as compared to (28a, b)) apple BEI all Lisi buy-away-LE 'The apples were all bought by Lisi.' Interestingly, the English get passive does not allow quantifiers to float in the same position either: (v) *They got all arrested. (Fleisher 2008) We simply acknowledge this as a general restriction, not peculiar to (ii) and (iv), but will not attempt to explain the reason for this restriction.
Control and raising passives, and why Mandarin does not smuggle 399 These differences can be explained if we assume that the English passive has smuggling while the Mandarin one does not. As shown in (31), after the movement of PartP to Spec, VoiceP, Voice' includes by, the Agent DP the police and the trace of PartP. That's why the by-DP behaves like a constituent. 9 However, in the Mandarin passive, as shown in (26-27), VP does not move to Spec, VoiceP but stays in situ within Voice'. Therefore, the Voice bei and the Agent DP do not form a constituent. 10 An apparent counter-example to the above explanation was suggested by Shi andHu (2005, p. 216), quoting Chen (2001), as shown in (33) In (33), bei-DP seems to be able to pass the coordination constituency test. However, this test should not be treated as the one for the constituent status of ''bei-DP''. Rather, it is a phenomenon of right node raising (RNR), as argued by Huang (1999) and Xiong (2010). It is similar to (34) which is a typical case of RNR: According to Wilder (1999, p. 2), in an RNR construction, if the shared constituent a surfaces the final conjunct, gap(s) corresponding to a must be at the right edge of their non-final conjuncts. In (34), the shared constituent the book is in the final conjunct Mary read and the gap that corresponds to it is at the right edge of nonfinal conjunct John bought. The ''coordinated'' bei-DPs in (33) are in the similar configuration, as shown in (35) As pointed out by a reviewer, since in English by and the police form a Voice' (not VoiceP), the fact that by-phrase can be preposed seems to suggest X'-movement. We doubt the status of an X'-movement but defer to Collins (2005), who rejects the stipulation made by Chomsky (1986) that intermediate categories cannot undergo movement, but did not go further to explain the issue. The relevant point is that if Collins is right about English by-phrases, then the Mandarin bei ? DP is not a phrase. 10 We thank Grant Goodall (p.c.) for a question that led us to this argument.
In (35), the shared constituent (the VP tunshi-le 'swallow-LE') surfaces with the final conjunct bei yi-ge guangmo de kongjian 'BEI a wild space', and its corresponding gap is at the right edge of the non-final conjuncts bei shu 'BEI trees' and bei cao 'BEI grass'. Therefore, (33) is a case of RNR. Furthermore, according to Postal (1974), Gazdar (1981), and Williams (1981) among others, the function of RNR is to identify the constituency status of the raised rightmost part (such as the book in (34)), but not that of the remnant (such as John bought or Mary read in (34)). Therefore, (33) has the RNR structure illustrated in (35), which does not establish the constituency of a bei-DP sequence.

Smuggling is not only unnecessary but also impossible in Mandarin passives
According to Soh (1998), verb raising is obligatory in Mandarin. This suggests, at first sight, the possibility (C. Collins, p.c.) that smuggling may still be applied after V moves to v, as in (36).
In (36), after V 'hit' raises to v, the VP containing the trace of V could undergo remnant movement to Spec, VoiceP, thus smuggling the internal argument Zhangsan across the external argument Lisi. However, we argue that this alternative must be excluded. Passivizing the Theme object in (37a), we get (37b). If there is VP movement to Spec, VoiceP, the dative PP 'to Lisi' should be able to move above bei-DP, contrary to fact, as in (37c). 11 11 Although our account, in terms of the availability of an inner Spec, vP, offers an explanation for why smuggling is not needed, nothing we have said so far explains why smuggling is not allowed in Mandarin. While we do not have a fully developed theory for this latter situation, we think a plausible explanation is available from the assumption, suggested to us by a reviewer, that vP is a phase in Mandarin passives (though not in English passives). Assuming Phase Theory, the derivation indicated in (36) is excluded by the Phase Impenetrability Condition (PIC, Chomsky 2000Chomsky , 2008 if vP is a phase. [VP movement through the inner Spec of vP is prevented by considerations of anti-locality (Bošković 2005;Grohmann 2003).] The phasehood of the vP under bei in Chinese may follow from the well known fact that it retains active morphology, unlike the English vP with a passive participial phrase. Other questions arise that go beyond (37)

Conclusion
In this paper we have shown that while English and Mandarin passives share similar underlying structures in observance of UTAH, they differ in their (non-)use of smuggling. We argued that this difference cannot be attributed to the supposition that Chinese passives involve only control or predication, but provided evidence for the possibility of a raising derivation involving A-movement. Instead, we tie the ability to do without smuggling in Mandarin to the existence of vP-internal movement, construed as part of a more general parameter that also applies to the CP domain, which provides the object with an intermediate landing site on its way to the subject position without violating minimality conditions. The conclusion that smuggling is not universal is not necessarily a bad result: While UTAH and minimality are presumably universal requirements, languages may employ different strategies to satisfy them. As we have shown, the non-universal view of smuggling allows us to tie together a number of otherwise unrelated differences among these languages-with respect to the constituency of the 'by phrase', the distribution of quantifier float and clause-internal objectpreposing. It should be noted that this paper has not invalidated Collins' smuggling analysis of English passives. Indeed, some of the correlated differences follow from the assumption that while smuggling does not occur in Mandarin, it does in English.