The default end result which one would be inclined to associate with a particular event depends on the participants in that event. Consider the set of sentences in w19.
John hammered the metal. (flat, into a ball) John hammered the porcelain. (to dust, cracked) John hammered the fruit. (to a pulp)
For each different kind of object which John hammers, a different default end result would be inferred. It is true that in each of the above examples that result is a state. It is therefore conceivable that all that should be encoded in the TELOS should be that the result is some state. In that case a pragmatic constraint would clearly have to be recruited to account for object-dependent variations in the construction of felicitous resultative sentences, as shown by the contrast between w19 and w20.
??John hammered the metal to dust. ??John hammered the porcelain to a pulp. ??John hammered the fruit flat.
Even simply encoding that the canonical result for hammer is a state would be too restrictive, in light of examples such as w21, in which the resultative phrase adds a change of location rather than a change of state.
John hammered the nail into the wall. John hammered the board out of the window/off of the slide/onto the ground.
Also consider the sentences in w18. w18a primarily specifies the creation of an object (a portrait of John) rather than the covering of a surface by paint, in contrast to John painted the wall, while w18b is a resultative which specifies not a change of state of the painted object (a star), but the location at which the stars are painted. Specifying that the canonical endpoint of the painting event is for the painted object to change state can't reflect the full intended meaning of such sentences.
I painted a portrait of John. I painted a star on each of my Christmas cards.
Thus the assumption of a canonical end state buys us little in accounting for the restrictedness of the resultative construction: no encoded canonical end state can be made specific enough to account for restrictions which arise from the use of the verb in a specific sentence, nor general enough to account for the full range of possible felicitous uses of the verb in the resultative construction. I will argue in Sections 4.4.3, 4.5, and 4.7.4 that the application of a pragmatic constraint to the construction better models the data.
Furthermore, Wechsler's and 's motivation for the representation of the canonical result in terms of potential telic interpretations of the verbs is suspect, since it depends on an analysis of aspect driven by lexical representations. Consider Wechsler's (, p. 6) discussion of the representation of the word paint:
Not all painting events have a definite end point, so the value for telos is optional (indicated by parentheses). But painting events can be telic; and if a painting event is telic, then the result state is canonically a state involving the state of the painted participant.Painting events can be both atelic and telic, as is clear from the two variants in w16. The first sentence has an atelic main clause (with no inherent endpoint) while the main clause is telic (with an inherent endpoint) in the second, according to the standard Vendlerian tests.
John painted the wall for two hours. John painted the wall in two hours.
The most salient interpretation for John painted the wall seems to be the telic one, but this generalisation does not hold across all verbs which can function as control resultatives. Consider the sentences in w17, with my grammatical judgements. The sentence in w17c is from (1996, p. 13).
The blacksmith/child hammered the metal. The blacksmith/child hammered the metal for three days. ?The blacksmith hammered the metal in three days. #The child hammered the metal in three days.
The telic interpretation of w17a is far more marked than the atelic interpretation, and the questionable felicity of the telic variants suggests that the representation of a generally available ``canonical end state'' for the activity of hammering is overly strong.
The question of how the telic interpretation of such sentences is arrived at must be carefully considered. It is by no means obvious that the (potential) result state should be represented in the lexical entries of these verbs. The aspectual category of a proposition can be affected by adverbials (Moens and Steedman 1988), and even by the kind of objects which serve as a verb's arguments (contrast John built a house in a day with *John built houses in a day (Verkuyl 1989, Krifka 1989). In short, to quote Moens and Steedman (1988, p. 17), ``aspectual profiles are properties of sentences used in a context: sense-meanings of sentences or verbs in isolation are usually compatible with several (or even all possible) Vendlerian profiles''.
It even seems that our world knowledge plays a role in determining the appropriate aspectual category for a proposition. Consider the sentences in w17 again. The difference in felicity between w17c and w17d is that for the former we are able to infer an end state for the metal and interpret the ``hammering the metal'' event as telic on the basis of our knowledge that that part of what a blacksmith does is to hammer metal into particular shapes, i.e. to make tools out of metal. In the case of a child, on the other hand, we don't have any knowledge of a ``standard'' end state which his/her hammering could lead to and so the sentence is odd. Pragmatic reasoning is again seen to interact with the interpretation process. The assumption of a lexically-encoded telic interpretation for a verb cannot adequately accommodate these pragmatic influences.
It could be argued that the independent motivation for the lexical representation of a canonical end state comes from the semantics of the verbs, that verbs like laugh can only appear as a raising resultative because they are never associated with any canonical end result while verbs like hammer and run can appear as control resultatives because they do have telic interpretations. However, the telic interpretation of hammer seems only to come from the fact that it can appear in a resultative construction. Similarly, the telic interpretation of manner of motion verbs like run arises only when it appears with telic prepositional phrases or telic arguments (e.g. run a mile). Thus it seems that the syntactic constructions in which a verb can appear determine the range of its interpretation. This means that the telic interpretation of these verbs stems from properties of a particular construction and properties of elements like PPs and arguments interacting with the core meaning (semantic relation) expressed by the verbs. It suggests that the syntactic context in which a verb is used can influence its potential interpretation, and that therefore the semantics of a verb in different constructions cannot be considered an independent criteria for determining the range of lexical entries it should have. The notion of canonical end state simply cannot be isolated from the syntactic constructions in which a telic interpretation is licensed.