[Warning: this is long and pretty technical -- but, I think, necessarily so.]
From Ben Zimmer on July 5, this wonderful relative clause example (with the crucial part boldfaced), from a NYT story about Gov. Rick Perry of Texas:
(1) Mr. Perry, whose aides say will make a decision within weeks, has been meeting around the country with potential fund-raisers… (link)
In terms that have become customary in talking about relative clauses (and other “extraction” constructions): the relative clause in (1) has a gap in its VP, a subject gap in the clause that is complement to say:
say [ ___ will make a decision within weeks ]
The gap is filled by the relative pronoun who in whose.
Framing this in somewhat more neutral terms (without reference to gaps, fillers, or extraction): who in whose serves as the subject of the VP will make a decision within weeks.
But at the same time, whose aides serves as the subject of the VP say will make a decision within weeks. So there are two syntactic-relation linkages here: the whole NP whose aides to the larger VP, and the relative pronoun who (within whose aides) to the smaller VP.
Ben judged (1) to be somewhat odd, despite its source, and I agreed with him, but he quickly came up with other parallel examples from equally respectable sources, so we concluded that the pattern of linkages in (1) is not to be labeled as generally ungrammatical in English (though there are speakers who prefer alternatives to it). It’s not clear how to analyze such double-linkage examples, but (as Geoff Pullum noted in correspondence with us) movement analyses, in which constituents are literally extracted from other constituents and moved to the front of the clause, would seem to offer no plausible source for them.
Before going on to further double-linkage examples, I’ll note two alternatives to double linkages that are less problematic analytically; instead of
(1) Mr. Perry, whose aides say ___ will make a decision within weeks, …
(with whose and a gap), a gapless alternative:
(1′) Mr. Perry, whose aides say he will make a decision within weeks, …
(with the pronoun he, anaphoric to Mr. Perry, instead of a gap), and a whose-less alternative:
(1″) Mr. Perry, who his aides say ___ will make a decision within weeks, …
(with the wh-pronoun who and the possessive pronoun his, anaphoric to Mr. Perry).
Now for some more double-linkage examples from Ben, two non-restrictive and one restrictive:
(2) Local shopkeepers are to start a “Come back Pingu” campaign with wooden cutouts of the bird, whose captors say will be returned in time for the village’s cherry festival in May. (link) [gapless alternative with it; whose-less alternative with who its]
(3) Strauss-Kahn, whose lawyers say will plead not guilty, faces arraignment today, meaning there’s a media circus at Manhattan Criminal Court. (link) [gapless alternative with he; whose-less alternative with who his]
(4) The Senate voted overwhelmingly on Tuesday to put new restrictions on the credit card industry, passing a bill whose backers say will make card-issuers spell out their terms in fewer words, using plain English, and treat customers more fairly. (link) [gapless alternative with it; whose-less alternative with which its]
Given (1)-(4), you can construct examples with other “gapped” pronouns:
Ms. Perry, whose aides say ___ will make a decision within weeks, … [gapless alternative with she; whose-less alternative with who her]
The Perrys, whose aides say ___ will make a decision within weeks, … [gapless alternative with they; whose-less alternative with who their]
… passing bills whose backers say ___ will improve things for customers … [gapless alternative with they; whose-less alternative with which their]
Some gapless examples collected by Ben Zimmer:
But uncertainties have been heightened by the weekend arrest of Strauss-Kahn, whose lawyers say he will deny charges of sexual assault. (link)
Felix, whose doctors say he will still be healthy enough for a transplant for a year or so, is believed to be the first person to miss out because of the policy that went into effect Oct. 1. (link)
Keith Bray, whose Florida lawyers say he will not yet discuss the matter, in a court filing in Tampa claims he deserves around 7.5% of the treasure that Odyssey showcased in plastic buckets filled with gold and silver coins earlier this year. (link)
There’s an important way in which the gapped and gapless examples aren’t fully parallel, but to show that, I need to say some basic things about relatvization constructions, starting with simple things and building up.
There are two quite different types of linkages involved in relativization: filler-gap linkages (which I suggested above should be thought of as syntactic-relation linkages between constituents) and antecedent-anaphor linkages. A simple example:
(a) Mr. Perry, who reporters have criticized ___
Here there’s a filler-gap linkage (or, in other terms, a linkage between the verb criticized and its direct object who), indicated by underlining; and a linkage between the constituent Mr. Perry as antecedent and the pronoun who as anaphor, indicated by boldfacing.
Then with a more distant filler-gap linkage:
(b) Mr. Perry, who the newspapers say ___ will lose the election
And with whose + Nom , where the filler and the relative anaphor aren’t coextensive:
(c) Mr. Perry, whose aides the newspapers have criticized ___
Then on to relativization on subjects, with
(d) Mr. Perry, who says no opponent could win
Here you might posit a gap in subject position, though the analytic status of such gaps is a matter of some technical wrangling; there’s an obvious sense in which who is simply the subject of the relative clause, period.
Then a combination of (c) and (d):
(e) Mr. Perry, whose aides say no opponent could win
Here, whose aides is the subject of say no opponent could win. Finally, a combination of (b) and (e):
(f) Mr. Perry, whose aides say ___ will win the election
This is the double-linkage construction; whose aides is the subject of say will win the election, and who is the subject of will win the election. The first of these syntactic-relation linkages is unproblematic, but the second is puzzling, since it seems to “go down into” the word whose to pick out the wh-relativizer in it, an option that, so far as I can see, isn’t available in anyone’s framework for filler-gap linkages.
One possibility is that this option should simply be stipulated, as a kind of footnote to general accounts of filler-gap dependencies. This idea is not as unpalatable as it might at first appear to be — a great many syntacticians just hate stipulations — since it would provide an easy way to allow for speakers (I suspect there are a fair number) who simply reject the doubly-linked examples.
Digression: on the gapless examples. In addition to the antecedent-anaphor linkage between the head of a relative clause and a wh-relativizer in it, any number of personal pronouns anaphoric to the head can occur within the relative clause — those boldfaced in the following examples, for instance:
(a’) Mr. Perry, who reporters have criticized ___ on account of his policies
(b’) Mr. Perry, who the newspapers say ___ will lose his election bid
(c’) Mr. Perry, whose aides the newspapers have criticized ___ because of their support for him; Mr. Perry, whose aides he has asked ___ for support
(d’) Mr. Perry, who says no opponent could win against him; Mr. Perry, who says he could defeat any opponent
(e’) Mr. Perry, whose aides say no opponent could defeat him; Mr. Perry, whose aides say he could defeat any opponent
And in fact in gapless examples like
(f’) Mr. Perry, whose aides say he will win the election
In all of these examples, the most easily available interpretations are the ones with the personal pronouns anaphoric to Mr. Perry (as indicated by the boldfacing), but the pronouns could pick up their referents from another antecedent earlier in the discourse. To see this, look at examples like:
(g) As for Hillary Clinton, Mr. Perry, whose aides say she will certainly win the election, is unaccountably undaunted.
(h) Barack Obama leads the field by a wide margin. Yet Mr. Perry, whose aides say he will certainly win the election, is unaccountably undaunted.
So the pairing of gapped and gapless examples I displayed earlier is misleading; the gapless examples simply have ordinary anaphoric pronouns, while the gapped examples exhibit a different sort of linkage.
The involvement of subject gaps. Ben Zimmer’s original examples all involved subject gaps; these are especially striking cases. But parallel examples are easily constructed with non-subject gaps:
Mr. Perry, whose aides will abandon ___
Mr. Perry, whose aides say they will abandon ___
Again, the gap-filler dependencies “go down into” the wh-relativizer whose, to pick out who as the filler of the gap.
The whose-less alternative. Let’s look back at the whose-less examples, in which the function of whose is performed by a sequence of a relative pronoun (who or which) and a possessive pronoun (especially his, her, its, or their); see the comments on (1)-(4) and similar gapped examples above. By splitting whose into two parts, these examples provide an entirely unproblematic source (in the first part) for the puzzling gap-filler dependency in the gapped examples. Schematically, we have clauses of the form:
[ 1:RelPron + [ [ 2:PossPron + 3:Nom ] + 4:VP ] ] ]
e.g., [ 1:who + [ [ 2:her + 3:aides ] + 4:say will … ] ] ]
Element 1 is then available as a filler to link to a gap within element 4 (as in say ___ will make a decision soon or say they will abandon ___). (Meanwhile, 2 + 3 is a subject NP with 4 as its VP.)
These observations could be turned into an analytic proposal, in which whose represents a fusion of two elements (1 and 2 in the schema above, a relative pronoun and a possessive pronoun). This would be a type of fusion different from the ones examined in CGEL — fused-head constructions (see discussion at the end of this posting) and fused relatives (traditionally called “free relative clauses”) — but at least the idea of fusion is out there.
Another possibility would be to say that the behavior of whose-relatives just has to be stipulated in the grammars of speakers who allow the doubly-linked examples, but that its historical origin lay in fusion, which has now become grammaticized for some speakers. (That would be not unlike what I’ve said about the wh-that construction in modern English, as in
I wonder how many people that were at the party.
which I’ve suggested originated in a syntactic blend of interrogative wh-complement clauses with relative clauses, but has become simply another complement construction for some speakers.)
Some support for the grammaticization account for whose-relatives comes from examples that do not submit easily to simple fusion, for instance:
Mr. Perry, most of whose aides say ___ will make a decision soon
(Admittedly, this is pretty far out for me, but then I wasn’t entirely comfortable with (1).) Here, whose can’t be straighforwardly decomposed into who his; instead, it would have to be something like who most of his, with who his split up by most of.
This is merely a progress report on several days of thinking and writing, inspired by Ben Zimmer’s query about his data and Geoff Pullum’s comments on them, not a polished analysis.