Noncompositional

(Quite a bunch of posts recently. I’m in the home stretch - basically just two squibs left, using mostly the same data. That data will be the topic of this post.)

According to one well-known account of argument structure (Jane Grimshaw’s 1990 monograph Argument Structure), some types of nouns have argument structure (nearly) identical in type to verbs. That is, semantic arguments in the lexical conceptual structure (”frames,” in some terminology, though they would be highly deficient frames) are linked to a separate representation that provides a link from the lexicon to the syntax. All elements of the argument structure must be realized (locally, I suppose) (except in certain constructional environments, like generics, instructional imperatives, so on and so forth). So is the case with nouns. And what sorts of nouns have argument structure? Those that denote “complex” (temporally/aspectually, I suppose, though she never says outright) events. She provides some examples that are questionable and probably would like a better analysis, but that’s for another time. Right now consider one event nominal (which may or may not be complex), use:

• The use of Martian technology to influence Terran politics is strictly prohibited.
• His use of new technology was impressive
.

Use shows some hallmark properties of being an arg-structure-having noun (obligatory complements, event control, and so on). But use has an interesting use as a sort of secondary predicate with the verb put:

• I believe that politicians are putting Martian technology to improper use

Roughly, the frame X put Y to use means that X causes some state-of-affairs such that someone uses Y. This someone must be under the control of X, who directed them to use Y to the advantage of X. So if management is finally putting the new computers to use, if they’re not using it, their subordinates (who they are “using”) are, to their benefit. On the other hand, if I convince my friends to use Red Hat Linux instead of Windows, then I did get them to use Linux, but I cannot be said to have put Linux to use. Just wanted to establish that “impliciture” (which is actually not an impliciture, but part of the semantics of put+use, or so I would claim).

Now, a question: is this the same use as in the first examples I gave? Certainly it denotes an event…maybe. Maybe it’s actually a state? Well, sort of, but it entails actual usage. Compare with put something in/into use or make available for use. In the latter, and possibly in the former, no actual use is entailed; a (strong) possibility is implicated. Okay, say it’s the same use: well, we’ve got a problem. Where are all the arguments? There’s no agent (put it to his use / use of the students), no patient (put computers to use of iBooks). Of course, the patient is elsewhere - it’s the object of put! And the agent…well, it’s related to the subject of put, but you can’t explicitly name the Agent. Seems to be a problem.

Another issue: consider my blog puts my girlfriend to sleep and the boss put him to work right away. Here the DOs of put are interpreted as the Agents of the secondary predicates, not the Patients. And come to think of it, there’s no way to interpret put X to use as ’cause X to use something.’ So there’s an asymmetry. This actually works fine, since in most theories of argument structure (all Generative ones, Chomskyan and otherwise (including CxG and probably HPSG)) the form of the external argument of verbs (and nominals) does not need to be specified by the verb/noun itself (and in Chomskyan theories, even the form is not specified). So the arg-struct already has an asymmetry that can be exploited. We can (in Minimalist jargon) first Merge the patient argument to use, and build the structure up from there. For unergative nouns, the external argument is Merged only later to “little n” (or “little v” if you later nominalize the whole sucker).

But this seems to miss something: in a case where the DO to put is animate, and the nominal is unergative, is the animate NP really acting as an Agent in the usual sense? Compare the sentences I made him work and I put him to work. Aside from the semantic differences in benefit to the ext. arg. of put that I mentioned earlier, maybe there’s a slight difference in the affectedness or level of agentivity of him from one case to the other? Actually, if anything the version with make seems to make the DO less agentive. Hmm… but if you compare it to I got him to work, the put case is much less agentive. On the other hand, compare I made her sleep and I put her to sleep. The former has a reading where she is agentively sleeping (cf. I made her go to sleep), but the latter doesn’t have that reading.

The evidence seems to be pointing in opposite directions. Okay…how about these two: A few brave soldiers put a large army to flight versus The commanding officer put his army to flight. Aha! The former says that the soldiers’ actions caused the army to take flight - though they would rather have been more brave and able to fight back. The latter sentence, however, is totally odd. It definitely cannot be used in the situation where the commanding officer orders his army to “Flee! Flee for the hills!” Similarly (?), you can scare birds, putting them to flight, but a pilot cannot *put his plane to flight (or at least his English teacher would look at him oddly). Actually, I’m not sure what to think of that. Is a flying plane an Agent or a Patient/Theme (man do these theta-role labels suck)? If what I’d like to think about put is true then put a plane to flight is bad because it’s like making something do something against its will, where a plane has no will per se. Eh, it’s a pretty lame explanation.

Okay, there’s even more to this: this is not a productive pattern. You cannot put someone to laugh, jog, or jump, or put books to sale. In fact, a more realistic (?) analysis from the start would be not to assume that we’re dealing with (a metaphorical extension of) put as a cause-motion verb, as I implicitly have been, but instead to call it a sort of light (or “support”) verb. But it is in fact a causative light verb, like give him a headache. As a light/support verb it can then be selected for by particular event-denoting nouns that cannot act as main predicators on their own. Thus put X to use would be similar to give testimony. Though the put does have its own semantics that other causative support verbs don’t necessarily have, so it’s really not all that light.

But this still doesn’t tell me why the arguments are suppressed in the put construction. Basically it means that you need to have null-type elements Merge at the beginning of the derivation (like PRO) which will then get their denotations from lexical items higher up in the tree. Or, you could move things around: but then there’s a problem: in the transitive case (use), you’d be moving something from the direct daughter to the lowest NP up to Spec,VP (the one headed by put, or LOC if you prefer); in the unergative case, you’d be moving something from Spec,nP to that same position. And why should they move? My latest theory is that, motivated by the affectedness evidence I talked about above, both the “Agent” and “Patient” roles are actually Patients, Merged directly with the nominals. With the transitive nouns, the Agent introduced by n is PRO, controlled by the ext. arg. of put; there is then movement from the lowest position up to Spec,VP. With unegrative nouns, the Patient is Merged with the noun, and then is moved first up to Spec,nP, and then again to Spec,VP — or alternatively, is actually PRO down below, bound by the Agent introduced by n. Of course I have no idea what motivates this movement: what motivates movement anyway? Features, yeah. What the heck features would I posit here? Beats me. And how would the speaker know to merge PRO anyway? Eh, the whole enterprise is not satisfactory.

So that’s where I am now. Except for more data, like put it to (a) (union) vote (as opposed to *put it to unions’ vote) and put him to a choice. Fun fun fun.