A causative construction is a sentence containing a transitive verb which essentially expresses ``cause to V-intransitive'', where V-intransitive is the intransitive counterpart of the verb (Levin 1993). Examples of this construction are in res36-res37, for which the causative (a) sentence entails the intransitive (b) one, (from Levin 1993:26-32).
John broke the glass. The glass broke.
The visitor rang the bell. The bell rang.
Particularly relevant data, involving verbs of manner of motion, are shown in res20-res5, (:188,(19)-(21)).
The soldiers marched (to the tents). The general marched the soldiers to the tents. ??The general marched the soldiers.
The horse jumped (over the fence). The rider jumped the horse over the fence. The rider jumped the horse. [directional phrase understood]
The mouse ran (through the maze). We ran the mouse through the maze. *We ran the mouse.
The tricycle rumbled across the sidewalk. They rumbled the tricycle across the sidewalk. *They rumbled the tricycle.
The examples suggest that a directional phrase must be present or at least understood when the manner of motion and sound emission verbs are used causatively. Thus the verbs differ with respect to their ability to undergo causativisation on their different senses, supporting the assertion that the lexical properties of these verbs change when they undergo the sense shift. Specifically, argue that the pattern in the causativisation data stems from the unaccusative argument structure of the verbs on a directed motion use. This conclusion depends on their view of causativisation, in which an external cause can only be introduced into an argument structure which has an empty external argument position. Since the causative form is felicitous on the directed motion sense of the verbs in res20-res5, the data suggest that on this sense the verbs have an unfilled external argument position -- that is, they have an unaccusative argument structure -- in contrast to the unshifted sense of the verbs which have an unergative argument structure and cannot appear in the causative construction.
The evidence from causativation, however, is weak at best. Levin and
Rappaport Hovav (1995:193) acknowledge that causativation
examples for the verbs of sound emission are infrequent, and the
attested examples they cite (such as res5b above) are not so
clearly acceptable.
Furthermore, most manner of motion verbs cannot
appear in the causative construction, even on a directed motion
interpretation, as shown in res6. The example res22b shows
great variability in acceptability when different objects and
directional phrases are chosen, as in res7, suggesting that the
causative use of run may be idiosyncratic or lexicalised with
highly specific selectional restrictions and a particular
interpretation varying from a true causative meaning. Other examples
of ``causative'' manner of motion verbs, as shown in res8, also
seem to have idiomatic, specific interpretations. In these cases, the
directional phrase seems to be optional as well, again suggesting a
lexicalised causative version of the verb rather than a causative form
derived
from an underlying unaccusative.
The child
limped/hobbled/ambled/meandered/swaggered/
sauntered/sashayed/wiggled
to the store.
#John
limped/hobbled/ambled/meandered/swaggered/
sauntered/sashayed/wiggled
the child to the store.
We ran the cat out of the house. *We ran the child to the store. *We ran the cat down the road. #We swam the mouse through the water maze.
We walked the dog (down the road). We walked Bill home. ?We walked the cat across the road. We jumped the horse (over the wall). ??We jumped the cat across the ditch.
The range of acceptability in this data points to the interaction of several knowledge sources in the evaluation of the felicity of instances of the causative construction. While there does appear to be a syntactic process which underlies the generation of causative forms of certain verbs, a purely syntactic treatment is bound to overgenerate and must be highly constrained through non-syntactic means. Constraints seem to derive both from semantics, e.g. from lexicalised (or conventionalised) meanings of particular verbs, such as walk and jump above, and from pragmatics -- the causative interpretation of a verb use must be made sense of in the context of use. We can in fact improve the felicity of certain instances of this construction by placing them in an appropriate context. Consider res39. This sentence, independent of a context, seems to make little sense. However, if we place it into the context of the discourse in res50 which establishes the causal influence John can have on the child limping to the finish line, it improves in felicity.
#John limped the child to the finish line. The children were racing towards the finish line. Sarah tripped over a treebranch and hurt her ankle. Determined not to see his favorite niece lose, John grabbed Sarah's hand, pulled her off the ground, and limped the child to the finish line. The process of causativisation, then, is only semi-productive. Its productivity is restricted via the mutual combination of word knowledge as dictated by lexical semantics and pragmatic inference. I will return to these issues to investigate the interaction of different knowledge sources in explaining such sentences, and also with respect to the resultative construction, in Section 4.3.1.
The lack of consistency in the causativisation data for verbs of manner of motion and verbs of sound emission means that sweeping claims about the underlying argument structure for these verbs on the directed motion sense are untenable on the basis of this data. This evidence is therefore inconclusive for the aim of justifying a particular syntactic encoding of the resultative construction.