ResP] which have the
associated meaning [NP cause NP
to be in Result State by NP V-ing]. This category
then leaves aside the directed motion use of verbs of manner of motion
and verbs of sound emission, as well as changes of state expressed
with unaccusative verbs. We will return to the treatment of these in
Section 4.7.2.
I propose to treat Resultatives in terms of a Construction Grammar construction, shown in resalt12. This construction captures the basic syntactic form of the relevant sentences. The basic semantics of these sentences is also reflected in this construction: NP causes NP to change to the result location or state expressed in ResP. The main verb expresses how the causation is achieved, and its meaning is integrated into the semantics of the construction as a whole via merging between PRED and CAUSE-EFFECT.
| Sem | CAUSE-EFFECT | < | cause | undergoer | goal | > |
| PRED | < | > | ||||
| Syn | V | SUBJ | OBJ |
OBL |
Motivation for the existence of a construction dictating the syntax and semantics of Resultatives comes from (a) the fixed syntactic nature of these sentences in violation of the structure dictated by normal verbal subcategorisation specifications and (b) the semantic similarities among sentences of this type, which cannot clearly be attributed to an interaction between the semantics of the participating components.
There seem to be no valid instances of this construction which vary from the syntactic pattern identified above (also argued by Jackendoff 1990). The position of the resultative phrase is fixed relative to adjuncts, as shown by resalt13-resalt14 (assuming readings in which the locational prepositional phrases are verbal modifiers rather than noun modifiers). A proposal like Wechsler (), in which the resultative phrase is treated on par with adjuncts and licensed by the same lexical rule, cannot account for this restriction on the relative order of the different types of phrases. Similarly, sentences such as res44a-b which correspond roughly to the resultative construction yet express that the subject rather than the object undergoes a change, are infelicitous. Sentence res44c in which the event which caused the death is phrasal (drank beer) is also not grammatical.
John hammered the metal flat in the workshop. *John hammered the metal in the workshop flat. *John hammered in the workshop the metal flat.
John laughed himself silly on Saturday evening. *John laughed himself on Saturday evening silly. *John laughed on Saturday evening himself silly.
*John played (cards) broke. *John drank beer to death. *John drank beer himself to death.
Furthermore, the appearance of unsubcategorised post-verbal elements
in this construction must be licensed, and in fact required, by some
mechanism. We saw above (Section 4.6.1) that the attempt
to explain this appearance purely in terms of syntactic constraints
and linking rules is inadequate. Goldberg (1995) and Wechsler
() both require these elements more or less by stipulation,
but the Goldberg account motivates it more convincingly in terms of
the semantic properties of the construction and within the context of
a general grammatical framework which has other instances of this
sort.
Thus I choose to follow Goldberg's approach.
The consistency of the interpretations assigned to this construction, despite the varying relations between the verbal head and the other constituents of the sentence as outlined in resalt1b, is striking. There are instances of each form of the construction -- varying in relations between the components and the type of resultative phrase -- which all convey essentially the same meaning. I summarise these in resalt20.
NP is a fake reflexive,
coindexed with the subject NP (V
intransitive)
John cried himself to sleep. [PP]
John caused himself to be asleep by John crying.
John laughed himself sore. [AP]
John caused himself to be sore by John laughing.
NP is the object of a transitive verb
John heated the water to boiling. [PP]
John caused the water to be boiling by John heating it.
John hammered the metal flat. [AP]
John caused the metal to be flat by John hammering it.
NP is neither subcategorised by
V nor a fake reflexive (V intransitive)
John sneezed the tissue off the table. [PP]
John caused the tissue to be off the table by John
sneezing.
John ran his Nikes threadbare. [AP]
John caused his Nikes to be threadbare by John running.
As pointed out in Section 4.3.1, Resultatives express causative changes of state. Levin and Rappaport Hovav (1995) suggest that the change in telicity from a process to an accomplishment which occurs in the interpretation of events expressed as Resultatives is sufficient to account for the element of causation. It is clearly insufficient, however, as suggested by the contrast with directed uses of the manner of motion verbs which are not interpreted as incorporating this element of causation despite a shift in telicity. The causation component of the interpretation therefore cannot be accounted for by any straightforward compositional mechanism, since the components of the Resultative construction are essentially the same as the directed manner of motion construction -- both constructions involve a result state or location being predicated of the referent of a noun phrase. The semantics of the resultative phrase does not directly integrate into the semantics of the remainder of the sentence: a meaning which does not come from either of these parts arises, seemingly without explanation. The explanation under the current proposal (following Goldberg 1995), then, is that this meaning comes from the semantics of the construction in which the words appear.
In proposing this construction, I have conflated two of the constructions which Goldberg (1995) puts forth. This single construction can be used to account for both her Causative Motion and Resultative constructions. This is because the difference between the two Goldberg constructions can be accounted for almost entirely by virtue of the kinds of resultative phrases which are used. If the resultative phrase expresses a location or locative path, then the Resultative is interpreted as a CAUSE-MOVE construction, while if the resultative phrase expresses a state or a path to a state, then the Resultative is interpreted as a CAUSE-BECOME construction. Furthermore, many verbs which can appear in one type of Resultative can appear in the other, as shown in resalt21.
John danced himself across the room. John danced himself to fame. John danced his feet sore.
The critics laughed the show out of town. John laughed himself out of a job. John laughed himself sore.
Those that cannot appear in both types can be ruled out on the basis of the pragmatic constraint alluded to in Sections 4.4.3 and 4.5 which requires a coherent causal relation to exist between the main event expressed in the sentence and the result state. So, for example, sentences such as resalt22 are infelicitous because it is difficult to imagine a context in which John hammering the metal or wiping the table could cause the metal or table, respectively, to move. It seems that if such a context could be found, the sentences should be acceptable, as suggested by resalt38 (Jackendoff, p.c.).
#John hammered the metal down the hall. #John wiped the table across the room.
John hammered the metal into the hose. John pounded the table down the hall.
These facts suggest that the CAUSE-MOVE and CAUSE-BECOME constructions are really just variations of the same CAUSE-EFFECT construction, and that the construction acquires a more specific interpretation through integration of the meaning of the result phrase. This integration occurs at the time of the merger of the semantics of the construction with the semantics of the main verb and the other components of the sentence. I will outline the process by which this occurs in the following section.
There are, in addition, thematic differences between Goldberg's Resultative and Caused Motion Constructions which I ignore in my conflation of the two. Goldberg treats the post-verbal NP as a theme in the Caused Motion Construction, but as a patient in the Resultative Construction. I treat the post-verbal NP as simply an undergoer in each case (note that this also follows from the Jackendoff (1990) account), (a) to bring the analysis in line with the representation proposed in Chapter 2, (b) to avoid well-known problems with identifying thematic roles, and (c) because the requirement that the post-verbal NP be a patient seems to be an instance of the pragmatic constraint of coherence of the causal relation rather than being an independent constraint. As evidence of the latter point, minimal variants of the examples in resalt5 shown in resalt23, in which the direct objects would certainly not be classified as patients under the standard analysis, seem to me to be pragmatically infelicitous rather than ungrammatical, in that they could be felicitous in a context establishing the appropriate causal relation.
#Sue watched the monster crazy.
on interp.: Sue caused the monster to be crazy by
watching it
#Sue loved the book tattered.
on
interp: Sue caused the book to become tattered by loving it