1 Pragmatic constraints on idiom syntax
In the domain of idiom syntax, Fanselow & Lenertová (Reference Fanselow and Lenertová2011) argue that structural constraints can explain reordering options of idioms. Consider the following examples (primary structural accent is marked by capitals; less strong structural accent is indicated by small caps):
Fanselow & Lenertová claim that once the structural accent on Korn has been assigned, the ordering die Flinte
${<}$
ins Korn has been created and cannot be changed in subsequent derivational steps. Szendrői (Reference Szendrői2017; henceforth KS) and Trotzke (Reference Trotzke2015a; henceforth AT) agree with this view, but AT argues that structural constraints are not the only factor restricting reordering patterns of idioms. Specifically, AT claims that in these structurally-driven approaches, ‘extra-linguistic (non-conventional) pragmatic factors may overwrite constraints ... and play a more crucial role in accounting for idiom flexibility’ (AT: 90) than previously assumed. Note that the marked word order in (1c) is not ungrammatical. It is merely unacceptable in the idiomatic reading; fronting ins Korn in the literal reading of the idiom (‘to throw a gun into a pile of grain’) is perfectly acceptable.
AT provides examples where the idiomatic reading is also available when structural ordering constraints like the one above are violated. One of these examples is discussed by KS:
In accordance with AT’s claims, KS points out that violation of the ordering zwei Fliegen
${<}$
mit einer Klappe in (2c) does not destroy the idiomatic reading. AT’s explanation of this observation refers to pragmatic considerations. In a nutshell, he claims that (2) might fall into one category with other cases he discusses in this context, for example, die Hand reichen ‘to help somebody’ (lit.: ‘to hand over a hand’). These idioms and (2) above share the feature that their literal readings denote quite ‘bizarre’ events (see also Trotzke Reference Trotzke and Stefanowitsch2015b for more cases and details on the pragmatics involved). AT argues that structural ordering constraints can be violated in these idiomatic cases because the competing literal reading is highly implausible (i.e. bizarre).
However, concerning the difference between (1) and (2), KS claims that ‘it is not obvious that the literal meaning is actually more accessible or plausible’ (KS: 226) in cases like (1). While I agree that ‘accessibility’ should be measured more accurately in future studies by experimental investigations, it is important to note that the literal reading of the idiom in (2) would correspond to an event where someone beats up two flies with one swat – an action that should be judged as bizarre across speakers. Under normal physical circumstances, it is not possible to beat up flies. It is only possible to kill flies by using a swat, and this is clearly expressed by German verbs like erschlagen, totschlagen ‘to beat a fly to death’ or treffen ‘to hit a fly’ – even a metonymic (i.e. non-literal) version of schlagen is an unusual (and, I suppose, highly infrequent) lexical choice in this context.
All in all, I assume that KS misunderstood this piece of data, and, given the other examples discussed by AT, I conclude that the general claim made by AT might be on the right track: structural constraints can be violated in cases where the literal reading of an idiom does not make much sense for pragmatic reasons.
2 Pragmatic constraints in the domain of island effects
KS also takes issue with AT’s pragmatic account of island effects. Concerning the general assumptions that AT formulates in this context, KS states that AT’s discussion ‘provides no motivation for these assumptions, or any evidence why they might hold’ (KS: 227). In the interest of space, I will confine my remarks to only one assumption that is mentioned by KS in this passage: ‘constituents conveying new information allow extraction most easily’.
One of the many observations AT cites in support of this claim are German data provided by Müller (Reference Müller2010: 68), who shows that was–für split out of external subjects improves by scrambling the object across the subject:
AT notes that scrambling the object den Fritz in (3b) alters the information structure such that the subject Bücher becomes ‘more focal’ (see AT: 99). AT shows that this is in accordance with prominent pragmatic accounts of island effects (Erteschik-Shir Reference Erteschik-Shir1973 et seq.).
Patterns like (3) at least constitute potential evidence for the general claim that ‘constituents conveying new information allow extraction most easily’. KS does not mention any of these data (see AT: 92–102 for more examples). Since she does not discuss the general claim stated above, it is thus not clear why KS rejects it.
