Idioms have frequently been used to argue for transformational accounts of certain syntactic phenomena. For instance, the fact that part of an idiom can undergo raising is often used to argue for a movement analysis of raising:
(1) The shit hit the fan.
(2) The shit seems to have hit the fan.
Same for promotion of objects to subjects in the passive:
(3) Someone spilled the beans. --> The beans were spilled.
(4) She pulled some strings (to get him hired). --> Some strings were pulled (to get him hired).
It has also been claimed that idioms argue against a transformational analysis of the by-phrase in passives, where it is related transformationally to the active subject, because no subject that is idiomatic in the active can appear in a by-phrase in the passive and retain its idiomatic meaning. A note on references first: This argument is probably a familiar one to most people, but I have had trouble locating a source for it. Bowers (2010, 11-12) cites Marantz 1984, pages 26-27, for the argument, but I cannot find the argument there, or anywhere in Marantz 1984. Postal 2004, page 255, states the argument explicitly as following from the view of the passive in Chomsky 1981, but does not credit the argument to anyone. So, I do not know where the argument originated, but everyone seems to be aware of it.
Now, to the argument itself. There are not very many idioms that include a fixed subject, verb, and object, but there are a few. None of them permit a passive with the fixed subject in a by-phrase:
(5) The shit hit the fan. --> *The fan was hit by the shit.
(6) That's a case of the pot calling the kettle black. --> *That's a case of the kettle being called black by the pot.
(7) Elvis has left the building. (="the event is over") --> *The building has been left by Elvis.
(8) The ram has touched the wall. (="it's too late to turn back now") --> *The wall has been touched by the ram.
The point of this blog post is that this argument and corresponding conclusion are correct, despite several recent claims to the contrary. Postal 2004 (pages 255-256) and Bowers 2010 (pages 12-13) claim that there are idiomatic subjects of actives that can appear in by-phrases in the passive. However, none of their examples are pertinent. Relevant examples have to involve idiomatic subjects, obviously. However, this by itself is not good enough. If it is just the subject that has an idiomatic (or just metaphorical) sense, then this is not good enough. An NP with a self-contained special interpretation should be able to appear in a variety of positions, independent of anything else. It is only when the relevant NP has its special interpretation only by virtue of other elements that it appears with that the example becomes relevant. For instance, in the shit hit the fan above, the shit has its special interpretation only when it occurs with the verb hit and the object the fan. "The shit smacked the fan" and "the shit hit the air conditioner" do not have the idiomatic meaning. That is why separating the shit from the verb hit in (2) shows something: the two pieces must occur together at some level of representation for the idiomatic reading to obtain, since they only have their idiomatic meanings by virtue of occurring together.
So now let us turn to the putative counterexamples. Here are the examples from Postal 2004, pages 255-256:
(9) The lovebug bit Ted. --> Ted was bitten by the lovebug.
(10) A little bird told me that. --> I was told that by a little bird.
(11) Birds of a feather may decide to flock together. --> It may be decided by birds of a feather to flock together.
(12) Old dogs may even decide to learn new tricks. --> It may even be decided by old dogs to learn new tricks.
The problem is that these are not true idiomatic phrases that include a fixed subject, verb, and object. What is idiomatic in all of Postal's examples is just the subject itself (and even that is not very idiomatic; it is usually just metaphoric). The verb and object have their literal meanings. In addition, the subject phrase can appear with various different predicates, and with different grammatical roles:
(13) the lovebug has struck Bill, Bill has the lovebug, Bill has the bug
(14) a little bird is broadcasting that, a little bird whispered that to me, I heard it from a little bird
(15) birds of a feather hang out together, they're birds of a feather
(16) I'm an old dog, you can't teach an old dog new tricks, old dogs can't learn new tricks, old dogs and new tricks don't mesh
These "idioms" do not have a fixed form and an unexpected meaning, like the shit hit the fan, but merely a metaphoric meaning: using the bug to refer to some kind of obsession or infatuation; a little bird to mean an anonymous source; and so on. So long as these phrases occur with words that are compatible with their use in this metaphor, such as passing on information in the case of a little bird, there is no barrier to their use. It is therefore not surprising that they can appear in a passive as in Postal's examples.
Just to shore this up with textual data (searches performed 1/14/2011), I find "you've given me the tennis bug" at http://www.mylondon2012.com/mascots/pictures/youve-given-me-the-tennis-bug/; "Perkins caught the tennis bug from her older brother" at http://www.tennisrecruiting.net/article.asp?id=1121; "Moving to Atlanta was just the spark she needed to ignite the `tennis bug'" at http://www.tennisdynamics.net/facilitymanagement/team/index.asp; and numerous other examples. Googling "heard it from a little bird" turns up 47,400 results; one includes "your little bird is nothing but a tattletale!" (Cheater's Guide to Speaking English Like a Native By Boye Lafayette De Mente, on Google Books.) Googling "they're birds of a feather" gets 28,100 results. Googling "I'm an old dog" gets 361,000 results; of the first ten, two clearly mean it in the sense of "you can't teach an old dog new tricks" (so not all of those 361,000 are relevant, but many are).
Bowers 2010, page 13, adds three other putative examples:
(17) I felt as if a ton of bricks had hit me. --> I felt as if I had been hit by a ton of bricks.
(18) They believe the devil drove him to it. --> They believe he was driven to it by the devil.
(19) Photography/Hip-hop/Syntax/etc. fever has swept the nation. --> The nation has been swept by photography/hip-hop/syntax/etc. fever.
Again, these are not fixed idioms at all:
(20) I felt as if a ton of bricks had crashed down on me; it hit him like a ton of bricks; he'll come down on you like a ton of bricks; (actual expression is "like a ton of bricks")
(21) the devil made him do it, the devil forced him to do it, the devil got into him
(22) she has hip-hop fever; hip-hop fever has gripped the nation; do you have hip-hop fever? (Watch David Letterman and you will hear many different versions of "X fever"; this is just like "the X bug" in Postal's examples in 13.)
Once again, the special interpretation, to the extent that there is any, is just a property of the NP. That NP can occur with a variety of lexical items and in a variety of syntactic positions. Hence, these examples are irrelevant. (Again, google searches turn up numerous examples of these NPs without the verbs that Bowers takes to be part of the "idioms.")
I have not been able to find any counterexamples to the claim that true idiomatic phrases cannot appear in a by-phrase in the passive. I therefore conclude that idioms reveal an important asymmetry between object promotion and subject demotion in the passive. A chunk of an idiom can promote from active object to subject in the passive, meaning that the passive subject is derived from or related to the active object; but a chunk of an idiom cannot demote from active subject to by-phrase in the passive, meaning that the passive by-phrase is not derived from or related to the active subject.
Furthermore, it should be noted that the exact same active subjects that cannot be demoted to a by-phrase in the examples above can undergo raising:
(23) The shit seems to have hit the fan.
(24) The pot appears to be calling the kettle black.
(25) Elvis seems to have left the building.
(26) The ram appears to have touched the wall.
So, it is not the case that these particular idioms cannot be manipulated by syntactic rules. Rather, there is no syntactic rule that relates the active subject to the by-phrase.
This conclusion is problematic for a whole host of analyses of the passive. It is most problematic for two recent approaches that treat by-phrases as being identical to the active subject, namely Collins 2005 and Bowers 2010. Those approaches would have to add ad-hoc constraints to rule out pieces of idioms in by-phrases.
Bowers, John (2010). Arguments as Relations. Cambridge, MA: MIT Press.
Chomsky, Noam (1981). Lectures on Government and Binding. Dordrecht: Foris.
Collins, Chris (2005). A Smuggling Approach to the Passive in English. Syntax 8: 81-120.
Marantz, Alec (1984). On the Nature of Grammatical Relations. Cambridge, MA: MIT Press.
Postal, Paul M. (2004). Skeptical Linguistic Essays. Oxford: Oxford University Press.