Does Wh -in-Situ License Parasitic Gaps?

1 Introduction

It has been assumed that wh-in-situ does not license parasitic gaps (PGs) (Engdahl 1983). Compare the following examples:

(1)

1. a. Which documenti did John file ei [without reading pgi]?

2. b. *Who filed which documenti [without reading pgi]?

Some researchers, such as Nissenbaum (1999) and Kim (2001), argue against this hypothesis and propose that wh-in-situ is potentially capable of licensing PGs. Kim specifically predicts that the wh-phrases in wh-in-situ languages can license PGs. But Kim also mentions that this prediction is empirically difficult to test, since those languages known to lack syntactic wh-movement, such as Chinese and Japanese, permit "empty resumptive pronouns" in argument positions, and it is thus difficult to determine whether an empty category in an argument position is a PG or an empty pronominal (Kim 2001:103). In this squib, I argue that Kim's prediction does not hold. In particular, I will show that (a) wh-in-situ in Chinese does not license PGs, and (b) PGs are indeed attested in Chinese, and they must be licensed by syntactic wh movement.

2 Wh-in-Situ in Chinese and Parasitic Gaps

First let us examine wh-in-situ in Chinese. Emonds (2001) points out that PGs fall out nicely on Chomsky's (1982) analysis as a consequence of the general theory of operator binding. That is, a c-commanding wh-operator binds the PG as a by-product of binding the wh-trace. But a question comes to mind when we look at Chinese. Tsai (1994) argues that wh-arguments in Chinese are in fact variables unselectively bound by the question operator Q in Spec,CP. If the PG is a by-product of wh-binding, then it should be licensed by the question operator Q as a by-product of the binding of the wh-phrase. This prediction, however, is not borne out.¹

(2)

1. a. *Laowang [zai huijian pg_i zhiqian] jiuLaowang at meet      before already
   kaichu-le shei_i?
   fire-PERF who
   'Who did Laowang fire before meeting?'

2. b. *Laowang [zai du-guo  pg_i zhihou] jiu  diudiao-le
   Laowang at read-EXP    after then throw-PERF
   sheme wenjian_i?
   what document
   'Which document did Laowang throw away right after
   reading?'

The ungrammaticality of (2a-b) is unexpected on Emonds's (2001) construal. (2a-b) thus indicate that wh-in-situ in Chinese does not license PGs. The licensing of PGs involves more than Ā-inding.

3 Syntactic Wh-Movement and Parasitic Gaps

Are there PGs in Chinese? The answer is yes. What is more, PGs in Chinese must be licensed by syntactic wh-movement. To support this claim, let us turn to topicalization in Chinese.

Some researchers notice that topicalization (and relativization) in Chinese may not always involve syntactic wh-movement (see, e.g., Li 2002). An example is given in (3b), where the object weiyu 'tuna'is topicalized out of the relative clause in the object position. Since the relative clause constitutes a syntactic island, the topic weiyu 'tuna' cannot have moved to the initial position of the sentence through syntactic movement.

(3)

1. a. Weiyu, Laowang xihuan.
   tuna   Laowang like
   'Tuna, Laowang likes.'

2. b. Weiyu_i, Laowang yu-guo [e_j xihuan e_i de] ren_j.
   tuna    Laowang meet-EXP   like      MOD person
   'Tuna, Laowang has met persons who like [it].'

A surprising fact, however, is that topicalization of wh-elements in Chinese always involves syntactic wh-movement, as island effects typically show up. Consider the following examples:²

(4)

1. a. Sheme yu, Laowang xihuan?
   what  fish Laowang like
   'What fish does Laowang like?'

2. b. *Sheme yui, Laowang yu-guo  [e_j xihuan e_i de] ren_j?
   what fish Laowang meet-_EXP   like MOD person
   'What fish is it such that Laowang met persons who
   like it?'

Moreover, if the wh-phrases in (2a) and (2b) are topicalized, both sentences become grammatical. The PGs are licensed therein.

(5)

1. a. Shei_i Laowang [zai huijian pg_i zhiqian] jiu
   who Laowang at meet      before   already
   kaichu-le e_i?
   fire-PERF
   'Which person is it who Laowang fired...