3

The basic rule I follow is that the case of the relativizer is determined by the role played within the relative clause by the entity it represents.

Give it to [OBJ whomever you admire __ most]. (You admire him most.)
Give it to [SUBJ whoever __ should have it]. (He should have it.)
Give it to [SUBJ whoever you think __ should have it]. (You think he should have it.)

But what about a 'raised' subject?

Give it to [?? ???ever you want __ to have it].

Does the raised form You want him to have it govern, or does the underlying role He HAVE it?

Does the language's growing indifference to persnickety punctilio direct me to whoever as the 'unmarked' or 'more natural' form?

Or is Great Mother English sending me the message that I'm pushing the boundaries of the language and ought to rewrite?

StoneyB on hiatus
  • 68,905
  • Yay! This is the kind of question I want to see more of here! Because I have no idea how to answer it, and so I'm going to learn a lot. Sadly, we only have a handful of true linguists on this site dedicated to "linguists, etymologists, and enthusiasts", so I'm afraid your audience will be small :( Calling Dr. Lawler! – Dan Bron Jun 14 '16 at 18:50
  • 2
    I too hope @JohnLawler responds; I imagine he will have something to say about cyclic/postcyclic transformations which will at least clarify my very feeble grasp of that gnarly topic. – StoneyB on hiatus Jun 14 '16 at 18:53
  • 2
    'Raising' is a matter of semantics. With to-infinitivals, the intervening noun or NP in a complex catenative construction like this always belongs syntactically in the matrix; it functions as matrix object, so objective case "whomever" must be the answer. The intervening noun is only 'raised' inasmuch as the verb it relates to syntactically is higher in the constituent structure than the one it relates to semantically. – BillJ Jun 14 '16 at 19:29
  • Raising is both syntax and semantics, like most things in grammar. As to which case to assign, that's a matter between the speaker and their confessor, because it depends on what theoretical church they attend. They might go to Case Grammar of Arc Pair Grammar, for instance. Me, I contribute to the church that says "Don't use whom at all", so there is really no case to assign. – John Lawler Jun 14 '16 at 20:27
  • 1
    I'm with @BillJ, even though I can't push the lizard in Alice in Wonderland out of my head. – JEL Jun 14 '16 at 20:55
  • The syntactic status of gap here has nothing whatsoever to do with the semantics of 'raising'. As an exercise in syntax, gap can only be the syntactic object of the matrix verb and hence the pronoun case would strictly speaking be objective. It can be optionally overridden with subjective "whoever" but that's not the point here. – BillJ Jun 16 '16 at 19:46
  • @BillJ That's begging the question, which is whether the subject of the clausal complement of want is in fact an object of want or is merely represented as an object. – StoneyB on hiatus Jun 16 '16 at 20:44
  • @StoneyB The subordinate clause is syntactically subjectless, like most non-finites, though it's clearly understood as "whomever". In a diagram, the missing subject would be represented by a gap coreferenced to the object of the matrix verb "want", which would also be a gap as your example shows. It's essentially no different to any other complex catenative construction apart from the complication caused by the 'fused' relative pronoun "whomever" (meaning "the person whom"). I'll see if I can knock out a tree diagram when I get a moment – BillJ Jun 16 '16 at 22:00
  • @BillJ Don't worry with the diagram, I get you. And don't worry about the -ever, either; it works the same with "Give it to him who you think deserves it", "Give to him whom you want to have it", "Give it to him whom you think deserving", right? – StoneyB on hiatus Jun 16 '16 at 22:22
  • Okay. I agree with your first two, but I think the third one is essentially the same as the first one, so it should be subjective "who". It comprises: Give it to x + You think (that) x is deserving, where x is subject of the embedded finite content clause – BillJ Jun 17 '16 at 09:06
  • @BillJ But that's just the point: we don't say "You think he to deserve it", we say "You think him to deserve it", just as we say "You want him to deserve it." Where do we draw the line? – StoneyB on hiatus Jun 17 '16 at 10:06
  • I thought you said earlier that you "got it"! Those first two are ungrammatical. In the 3rd one, the intervening noun "him" is object of "want", but only the understood (i.e. semantic) subject of to deserve it. – BillJ Jun 17 '16 at 10:17
  • You have to take into account that in your last-but-one message, only the 2nd example is a complex catenative with an intervening noun. The other two were just relative clauses with additional embedded content clauses (or that-clauses if you prefer). – BillJ Jun 17 '16 at 10:21

1 Answers1

1

OK, I'll bite. In fact, I think the most reasonable position -- and the one adopted in Relational and Arc Pair grammars -- is that a raised object is in fact an object -- a direct object, in fact. So it can passivize, for instance, (with some verbs, at least):

  • The detective believes that he has it. (no infinitive, no Raising)
  • The detective believes him to have it. (B-Raising to object)
  • He is believed to have it (by the detective). (B-Raising plus Passive of Raised object)

In relational grammars, B-Raising promotes a 1 in a lower clause to a 2 in a higher clause; and Passive promotes a 2 to a 1 in the same clause. Both of these occur on the same (higher) cycle.

This isn't a matter of "assigning case", however; the labial nansal infinx on whomever doesn't actually get "assigned" until some milliseconds before the speaker's lips and velic flap coordinate
to produce the /m/.

1's, 2's, and 3's in relational grammars are the basis of the system, and not just a tag that has to be assigned by the grammar. Rather, they refer to the categories Subject, Direct Object, and Indirect Object, respectively. These are the givens in RG, not derived or assigned categories. On the other hand, a lot of the givens in generative grammars aren't needed in RG; so it comes in handy.

John Lawler
  • 107,887