Noted by Wilson Gray on ADS-L on Monday, from his reading on Facebook. Wilson commented:
Remember the days of yore when people wrote: “depriving millions of health-care”?
The implicit analysis here is that the ordinary argument structure (hereafter, argstr) for the verb deprive has a Direct Object referring to a POSSESSOR in an act of deprivation, and an Oblique Object (marked by the P of) referring to a POSSESSION in this act. In abbreviated form: deprive has the argstr:
(1) SU: AGENT, DO: POSSESSOR, OO(of): POSSESSION
with the semantics that AGENT causes POSSESSOR to come to no longer have POSSESSION.
But the Facebook sentence has an argstr with a Direct Object referring to a POSSESSION and an Oblique Object (marked by the P for) referring to a POSSESSOR:
(2) SU: AGENT, DO: POSSESSION, OO(for): POSSESSOR
with the same semantics as in (1).
Now, alternative argstrs for the same verb are very common; the question is which verbs have which structures. Wilson’s judgment (which I share) is that deprive is fine in structure (1) — deprive millions of health-care — but not in structure (2) — deprive health-care for millions. (Divest is similar to deprive here.)
All of this is extraordinarily verb-specific. Eliminate is fine in (2) — eliminate health-care for millions — but not in (1) — eliminate millions of health-care (with millions understood as ‘millions of people’). Or in (3) — eliminate health-care from millions:
(3) SU: AGENT, DO: POSSESSION, OO(from): POSSESSOR
(Deny and reject similarly work in (2), but not, for me, in either (1) or (3).)
Deprive and divest are also out in (3). On the other hand, in my judgment, remove is fine in (3) — remove health-care from millions — not bad in (2) — remove health-care for millions — but wretched in (1) — remove millions of health-care. And then strip is fine all around: strip millions of health-care, strip health-care for millions, strip health-care from millions.
(Note on (2). Things are somewhat confounded by the existence of VPs of the form deprive X for Z, with an understood OO of Y, and Z an adjunct, not an argument, one denoting a period of time or a goal or purpose: I deprived myself (of sth.) for many weeks / for a good cause. This fact makes it very hard indeed to search databases for relevant examples of deprive in argstr (2),)
Already we are backing into another important fact about verbs in these argstrs: there is considerable variation from person to person. If you would not yourself say deprive healthcare for millions, your first instinct will be to suppose that it’s just an inadvertent error, possibly a blend of some kind — but that instinct would almost surely be wrong: you can expect that people who use deprive in argstr (2) will do so repeatedly, without feeling any need to correct themselves; for them, things like deprive healthcare for millions are simply acceptable, part of their variety of English. Maybe not yours, but theirs.
The word of argstrs is rife with variability. In the “dative alternations”, for example (V OBJECT to RECIPIENT ~ V RECIPIENT OBJECT). And in some others I’ve discussed on this blog: rid X of Y (rid the house of cockroaches) is preferred to rid Y from X (rid cockroaches from the house), though both are well attested; wash Y from X (wash stains from my shirt) is preferred to wash X of Y (wash my shirt of stains), though again both are attested.
From my 6/1/15 posting “This week’s diathesis alternation” (diathesis alternation is another name for an alternation in argstrs), on the case of bilk MONEY from VICTIM vs. bilk VICTIM of MONEY:
the usage literature on diathesis alternations is littered with peevish judgments, usually based on the fact that one alternant is more recent than the other; the implicit position is that we already have a syntax for the verb in question, so why should we add another? Sometimes speakers are accused of being “confused” about how to use the verb.
But as I have often pointed out on this blog, there are good reasons for creating new argument structures for existing verbs. For one thing, the new structures provide a way of marking arguments as discourse-new vs. discourse-old (Old Before New is an important factor in structuring discourse). For another, the difference between a direct object and an oblique object is often significant: the choice of DO can implicate a closer relationship between a situation and the relevant participant than the choice of OO, so alternants can differ pragmatically.
The larger point is that there is constant pressure to expand the range of argstrs available for particular verbs, because different argstrs do different work in discourse and conversation.