The purpose of this paper is to argue that the two causatives with have, that is, 'have + object + bare infinitive' as in (1a) and 'have + object + past participle' as in (2a), are represented on the semantic level as in (lb) and (2b), respectively:
(1) a. John had Bill wash the car.
b. [_<Event> CAUSE([_<Thing> JOHN]_<i>, [_<Event> GO([_<Event> BILL WASH THE CAR], [_<Path> TO([_<Thing> JOHN]_<i>)])])])]
(2) a. John had the car washed.
b. [_<Event> CAUSE([_<Thing> JOHN]_<i>, [_<Event> GO([_<Event> THE CAR BE WASHED], [_<Path> TO([_<Thing> JOHN]_<i>)])])])]
Aside from the kind of event involved in the complement, these two structures are isomophic in that the subject NP plays not only the role of 'Causer' but that of the Goal to which the event of the complement moves. Compare these structures with those of cause and make, which were proposed in Inoue(1989), as in (3) and (4):
(3) a. John caused Bill to work.
b. [_<Even> CAUSE([Thing JOHN], [_<Event> GO([_<Thing> BILL]_<i>, [_<Path> TO([_<Event> BILL_<i> WORK])])])]
(4) a. John made Bill work.
b. [_<Event> CAUSE([_<Thing> JOHN], [_<Event> BILL WORK])]
(3b) differs from (1b) and (2b) in that in the former 'Bill' is the Theme and the Event [BILL WORK] the Goal, while the opposite is the case in the latter; (4b) differs from (1b) and (2b) in that there is no thematic relation in the former.
Section 1 is devoted to a brief survey of the framework upon which the present analysis is based. Section 2 is concerned with how to distinguish have-constructions on semantic grounds. By using various types of test such as 'what happened was ...', 'what x did ...' and 'what happened to x ...', distinctions are made between statal have-constructions, the so-called "have-passive" and have-causatives. Section 3 presents evidence that argues for the structures (1b) and (2b). The main focus is placed on the issue of the subject NP as Goal by comparing have with make in respect of semantic behavior.
It follows from the discussion that if the present analysis is valid, it will enable us to account for (i) why the past participle is more likely to occur in the complement of have than that of make, and (ii) the striking parallelism that holds between have-causatives and other have-constructions.