Conventional combinations in pockets of productivity: English resultatives and Dutch ditransitives expressing excess

International audienceGoogle hits for “sneezed the napkin off the table” run into the thousands. The sequence of words has become common good, to the extent that we see it quoted with a range of different subjects (Adele(!), Alex, Bob, Donna, Frank, Fred, Jack, Joan, Joe, John, Mary, Paul, Pat, Rachel, Sally, Sue, Tom, I, He, She, The baby…) and often without reference to Goldberg’s (1995) original sneeze example. The example is captivating to anyone who first hears or reads it, because in its simplicity it manages to capture the essence of the constructionist movement, in which argument structure constructions still adopt a central place. In true constructionist spirit, one might even consider [X sneezed the napkin off the table] a construction all by itself, with (i) an open slot preferentially filled by a mono- or bisyllabic, somewhat old-fashioned, all-American proper name, (ii) a specific genre restriction (viz. academic writing related to linguistics or cognitive science) and (iii) a conventionalized interpretation (‘a verb may be plugged in larger syntactic frames which provide arguments not directly associated with the verb itself’). On a more serious note, however, this key example in present-day linguistics may have given rise to the idea that, provided there are no semantic clashes between word-level lexical constructions and the more schematic phrasal constructions which provide slots to them, ‘anything goes’ in grammar. And Goldberg (2003: 221) does little to nuance such a conception when she writes that “[c]onstructions can be combined freely to form actual expressions as long as they are not in conflict”.This paper argues against this view, extending some of the corrective perspectives advanced by Boas (2003), Iwata (2008) and Kay (2012) and drawing on corpus-based and web-collected data about English and Dutch intensifying argument structure constructions. My argumentation is rather complicated, which is why I give a preview of the different steps here. In Section 2, I will discuss the caused-motion pattern in English (of which the now-famous sneeze example is an illustration), claiming that it is idiosyncratically constrained and that novel uses may be analysed as the result of analogical extensions from conventional three-argument verbs and hence do not prove that there is a productive, Goldbergian (i.e. maximally schematic) caused-motion construction. In Section 3, I will turn to a pattern in English which does seem to be very productive, namely the Body Part Off Construction (henceforth BPOC, e.g. work one’s head off). Here, I will argue that instances of this pattern are not understood by means of a pragmatic reasoning process which ought to allow us to infer that the literal scenario cannot be meant by the speaker. This pattern may share the syntax of the caused-motion pattern (and, more generally, of the resultative pattern), but its semantics is action-intensifying rather than resultative. It is therefore a distinct pattern and cannot be served to prove the productive nature of the caused-motion pattern, let alone of the even more general resultative pattern. Speakers of English have to ‘learn’ the BPOC as a form-meaning pattern in its own right. In Section 3, I will use data from Dutch to support my claim that instances of the English BPOC are not formed or interpreted pragmatically, i.e. on the basis of the caused-motion pattern and general expressive or reasoning skills. If they were, we would have to find formally similar instances in Dutch. What we find in this neighbouring language, however, is a set of syntactically different patterns expressing excess, including a ditransitive one. In Section 4, I will take a more in-depth look at the BPOC, showing that it may best be analysed in terms of high-frequency learned instances and some creative extensions from these. In Section 5, I will show that the same kind of analysis should be adopted for Dutch intensifying double-object cases. Section 6 provides some contrastive observations about the two patterns, revealing more conventionality. Section 7 discusses the findings in the light of the tension between stored linguistic information and free application of constructional templates. Section 8 sums up the main points