The pre-encoding becomes clear upon examination of Wechsler's detailed illustration of the behaviour of the verb run within his approach (Wechsler , Appendix, p. 14-16). In this analysis the sentence We ran our Nikes threadbare is analysed using a variant of the lexical entry for run which does not have a telos value specified, which then leads to the addition of the non-argument NP to the verb's subcat list via the Raising Rule. The sentence John ran into a tavern, in comparison, is analyzed with a lexical entry for run for which the telos value is specified as a location. So we have one verb which can appear in two different resultative constructions, and as a result we have two variants of its lexical entry (cf. Jackendoff 1990, ; Goldberg 1995). A verb like laugh, on the other hand, is only ever used as a raising resultative and hence only has one lexical entry, with no telos value specified. The absence of an additional entry for laugh with a value for the telos feature is precluded only on the basis of the lack of examples of this verb in the control resultative construction.
World knowledge of situations involving the relations expressed by particular verbs could be considered an independent basis for encoding a canonical result, but we saw above with reference to the examples w19 and w20 that these `canonical' results can be highly context-specific. This will either lead to generation of individual lexical entries for a verb corresponding to distinct contexts, which is heavily redundant and fails to capture any generalisations, or specification of vague canonical results which will result in overgeneration. I therefore see no adequate justification for having both raising resultative and control resultative lexical entries for a particular verb, independent of that verb's behaviour. This pre-encoding undermines the explanatory adequacy of Wechsler's theory.