The End Result(ative)

This volume’s discussion notes include comments by Hans Boas and by Stephen Wechsler on our article, ‘The English resultative as a family of constructions’ (Goldberg & Jackendoff 2004), which we hope will raise further discussion about the nature of the semantic constraints and narrow productivity of the construction.¹

The main objectives of our article were to argue against the position that there exists a single uniform, general resultative construction, to detail the many ways in which semantics determines syntax, and to make clear that it is still necessary to state something special about the syntax. Boas and Wechsler accept most of our overall constructionist view but take issue with some of the finer points of our analysis of resultatives: the end-of-scale constraint on resultative adjectives and how exactly to predict the delicate balance between idiosyncrasy and generalization.

Wechsler (2005a) critiques a single paragraph in our article in which we mention that adjectives that conventionally appear in the resultative construction tend to be nongradable, coding a clearly delimited state (Goldberg & Jackendoff 2004:560). This description was an abbreviated version of a constraint discussed in Goldberg 1995: 195–97 under the heading ‘the end-of-scale constraint’ (195), foreshadowing Wechsler’s reference to a ‘maximal endpoint’ generalization. Wechsler’s discussion of maximal vs. minimal endpoint closed-scale gradable adjectives adds some welcome detail to our account.² Boas also amplifies our view by producing more cases of resultatives that are licensed by virtue of world knowledge, where no purely grammatical account will do the job. Indeed, much nuanced and interesting work has been done on the semantics of adjectives that we did not attempt to review in our brief allusion to the topic (including Boas’s (2000) detailed discussion).

While Wechsler appears to advocate predicting the occurrence of resultative adjectives on the basis of their inherent lexical semantics, Boas, along with Goldberg & Jackendoff 2004 and Goldberg 1995, observes that there is far too much idiosyncrasy to account for the full range of facts on the basis of general semantic principles. For example, it is not clear to us that more refined semantic analysis or more invocation of world knowledge is going to lead to an explanation of contrasts like our example: She ate herself sick/*She ate herself ill. (Goldberg & Jackendoff 2004:561, ex. 86). It is for such contrasts that we feel driven to invoke idiosyncratic semiproductivity. Boas’s discussion note centers around developing a detailed description of this partial productivity, a topic that we explicitly set aside in our discussion of the resultative, except to observe that the construction is clearly partially, but not fully, productive. Boas suggests adopting the fairly traditional approach of lexical subcategorization. In particular, he proposes that all verb/resultative pairs that appear in the British National Corpus (BNC) are understood to involve projections from distinct verb senses, even if they appear only once. Only examples that do not appear in the corpus, such as sneezing the foam off the cappuccino, are acknowledged by Boas to be truly novel, and he attributes these to one-shot analogies. The fact that the BNC is all of 100 million words underscores the oddity of treating each and every resultative that occurs in it as ‘conventional’; examples such as boil dry, seducing to death, stack flat, spoon full, and shitting silly would require special conventionalized senses of boil, seduce, stack, spoon, and so on (see Goldberg & Jackendoff 2004:256).

Our own position strikes a middle ground. We acknowledge a great deal of idiosyncrasy, while at the same time recognizing the existence of semantic generalizations. It is clear that much work is still required to thoroughly describe the partial productivity of the resultative and other constructions (Goldberg & Jackendoff 2004:563), much less explain how it comes about in the course of acquisition. Moreover, as we observe, the path resultatives (Bill rolled the ball down the hill) are far more productive than the property resultatives, the case with which Wechsler (2005a) and Boas (2005) are primarily concerned. And while property resultatives lie towards the idiosyncratic end of the continuum of productivity observed in argument structure...