The ECM Complement with Empty HAVE for “Order” and “Give”  

Yoshiki Ogawa and Tsuyoshi Sugawara

It has usually been assumed that the V-NP-to-VP construction headed by the verb order and the V-NP1-NP2 construction headed by the verb give are independent and unrelated constructions, and, as for the verb give , the V-NP1-NP2 construction and the V-NP2-to-NP1 construction are not transformationally related to each other.

In this talk, we will make three proposals about the syntactic structure and argument structure of order and give so that the above-mentioned three constructions are syntactically related to each other. Focusing on order , we make the first proposal that its indirect object, such as John in We ordered John to leave , which is normally analyzed as an element of the matrix clause that controls PRO in the infinitive complement, should be reanalyzed as the subject of the ECM complement whose verbal head is the invisible HAVE. This proposal entails that order is a dyadic verb with the argument structure of <Agent , Event> rather than a triadic verb with the argument structure of <Agent , Goal, Event>. This proposal is motivated by a couple of syntactic tests whose results distinguish order from persuade , a typical object control verb. The postulation of HAVE entails that the infinitival complement of order is [NP HAVE to do Y], which syntactically supports Nakamura's (1999) proposal that order selects an ECM complement with the imperative modality.

Adopting the spirit of Freeze's (1992) proposal that have is an amalgamation of copula be and a locative adposition, we make the second proposal that the dative-Case-marking adposition TO can be excorporated from HAVE to the head of the clausal complement of order which we tentatively call XP. When the first and second proposals are coupled with the descriptive generalization that the head of the clausal complement of a derived nominal cannot be phonetically empty (cf. Stowell 1981, Kayne 1984, Pesetsky 1995, Ogawa 2001), we can account for why the NP in the V-NP-to-VP construction is marked for dative rather than genitive Case when the verb order is nominalized, as in Sue's order to Harry to get out of the room .

Our third proposal is to extend the two proposals we have just outlined with the verb order to the verb give . This proposal entails that give is also a dyadic verb with the argument structure of < Agent , Event> rather than a triadic verb with the argument structure of < Agent , Goal, Theme>. A more specific proposal is that give that occurs in the V-NP1-NP2 construction also takes the ECM complement (of the category XP) whose verbal head is invisible HAVE, whose subject is NP1, and whose object is NP2. On the basis of this proposal, we will argue that the V-NP2-to-NP1 construction is transformationally derived from the V-NP1-NP2 construction by the optional excorporation of TO from HAVE to X. This argument will be supported by (i) the fact that the V-NP2-to-NP1 construction, but not the V-NP1-NP2 construction, can be nominalized; (ii) the fact that, with the Japanese verb ataeru ‘give', NP1 can be the antecedent of the subject-oriented anaphor zibun contained in NP2; and (iii) a new observation about the NPI licensing by deny , the negative counterpart of give .