Linguistic Inquiry

1 Introduction

Crosslinguistically, it is often observed that complementizers are optionally elided in finite complement clauses that are selected by a certain set of verbs. It has been a common assumption that the elided complementizer is an empty category lacking phonetic realization (see, e.g., Stowell 1981a,b, Pesetsky 1995, Bošković and Lasnik 2003), although the exact formulation of this assumption varies from one [End Page 339] proposal to another. However, Bošković (1997) has advanced another analysis, based on certain minimalist assumptions, which holds that no complementizer (or projection) exists in a complementizerless embedded clause (see also Doherty 1997).

Given these totally different views on complementizerless clauses, a theoretical issue arises: whether or not it is adequate to posit the existence of any phonetically null entity to deal with comple-mentizer deletion. In this squib, I argue, in the light of data from a dialect of Japanese, that a null complementizer, even though it lacks phonological realization, must exist as a syntactic entity in the CP projection in complementizerless subordinate clauses, despite Bošković's (1997) argument to the contrary.

2 Complementizer Deletion in a Dialect of Japanese

Traditionally, it has been assumed that a phonetically unrealized com-plementizer exists in the C-head position of a complementizerless subordinate clause. Since neither movement nor an anaphoric process that motivates the presence of a syntactic entity in the gap is involved in complementizer deletion, however, we cannot simply assume that there is an invisible category in the complementizer position. In addition, a null complementizer apparently does not contribute to semantic interpretation. Given the principle of economy of representation, which states that a clause can only contain as much material as is necessary to satisfy lexical requirements, it could be argued, as Bošković (1997) does, that in a complementizerless subordinate clause, no CP projection exists, let alone a phonetically null complementizer (see Bowers 1987).

This raises a theoretical issue: should a null complementizer exist in a complementizerless embedded clause, or not? Note here that Bošković (1997) does not deny the existence of null complementizers in general. In fact, on the assumption that declaratives are a default interpretation (i.e., no complementizer is needed to determine a declarative interpretation), Bošković claims that a complementizer is literally missing in the declarative embedded clause in John thinks Mary left , since its presence is neither motivated semantically nor required lexically.¹

The issue of whether or not an embedded declarative clause with no overt complementizer contains a covert complementizer can be resolved by looking at the behavior of adverbial particles in Japanese. In what follows, drawing on data from a western dialect of Japanese, the so-called Kansai dialect, I suggest, contrary to Bo'ković (1997), that a null complementizer must exist as a syntactic category in a complementizerless embedded clause (with a declarative interpretation), even if it is not phonetically realized. [End Page 340]

Let us begin by noting that Standard Japanese does not allow complementizers to be omitted in subordinate clauses. Nevertheless, Saito (1986) reports that in the Kansai dialect of Japanese, comple-mentizer deletion does take place in finite complement clauses selected by a certain class of verbs such as yuu 'say' and omou 'think'.²

(1) John-wa [Mary-ga ki-ta         (tte)] yuu-ta/omoo-ta.
John-TOP Mary-NOM come-PAST that say-PAST/think-PAST
'John said/thought (that) Mary came.'

Complementizer deletion in the Kansai dialect closely parallels that found in English. For one thing, manner-of-speaking verbs in English do not allow complementizer deletion in their complement clauses (see, e.g., Zwicky 1971, Stowell 1981a,b, Pesetsky 1995).

(2)

1. a. John thought (that) Mary was honest.

2. b. John shouted ?*(that) she was very hungry.

Likewise, in the Kansai dialect of Japanese, complementizer deletion does not apply in complement clauses selected by manner-of-speaking verbs.

(3) John-wa [kuruma-ga ki-ta         *(tte)] saken-dot-ta.
John-TOP car-NOM come-PAST     that shout-be-PAST
'John was shouting that a car came.'

For another thing, in English, when a complement clause is dislocated from its original position, deletion of the...